Embedded Computing Design

RSC# @ www.embedded-computing.com/rsc RSC# @ www.embedded-computing.com/rsc www.embedded-computing.com Volume 4 • Number 5 J u lY 2 0 0 6

COLUMNS FEATURES 8 Editor’s Foreword SPECIAL: Building trustworthy, survivable systems Embedded technology: Good for what ails you 18 Secure from the ground up By Jerry Gipper By Dan Dodge, QNX Software Systems 11 Embedded Perspective All these things that I’ve ... run TECHNOLOGY: Machines and humans meet – By Don Dingee audio, video, input, imaging, and recognition 12 Embedded Technology in Europe Embedded medical and biological applications 21 Mobile phone security – in your face By Hermann Strass By Seiji Inoue, Oki Electric 14 Eclipse Perspective and News Building embedded software with the APPLICATION: Medical – surgery, monitoring, Eclipse Process Framework and diagnosis By Chris Sibbald and Kurt Sand 26 Integrating secure, standards-based wireless connectivity for networked medical applications DEPARTMENTS By Shahin Hatamian, silex technology america 37 Editor’s Choice Products 30 Model-driven development for medical devices By Jerry Gipper By Irv Badr, Telelogic 44 New Products By Chad Lumsden PCI Express: Backplane implementations 39 Multi-CPU architectures with Advanced Switching extensions to PCI Express EVENTS By Stephen Christo, StarGen Freescale Technology Forum July 24-27 • JW Marriott and Ritz-Carlton Conference Center Orlando, FL www.freescale.com/ftf E-CASTS Upcoming: VoIP Developer Conference RapidIO August 8-10 • Hyatt Regency, Santa Clara, CA COM Express www.tmcnet.com/voipdeveloper/ DSP/FPGA Reconfigurable Computing www.opensystems-publishing.com/ecast

E-LETTER OpenSystems www.embedded-computing.com/eletter Advanced fuel dispenser brings in-store conveniences to driver’s seat COVER Publishing™ By Dan Harrell, Dresser Wayne Powered by state-of-the-art robotic technology, the da Vinci System seamlessly translates the surgeon’s hand movements into the more precise movements of the EndoWrist instruments, enabling surgeons to performOpenSystem minimally invasives procedures involving complex dissectionPublishing™ or reconstruction. WEB RESOURCES © 2006 Intuitive Surgical, Inc. Subscribe to the magazine or E-letter: www.opensystems-publishing.com/subscriptions Industry news: Published by: OpenSystems Read: www.embedded-computing.com/news OpenSystems Submit: www.opensystems-publishing.com/news/submit Publishing™Publishing™ Submit new products: © 2006 Embedded Computing Design www.opensystems-publishing.com/vendors/submissions/np All registered brands and trademarks within Embedded Computing Design are property of their respective owners.

/ July 2006 Embedded Computing Design

A n O p e n S y s t e m s P u b l i c a t i o n

Embedded and Test & Analysis Group n Embedded Computing Design n Embedded Computing Design E-letter n Embedded Computing Design Resource Guide n Industrial Embedded Systems n Industrial Embedded Systems E-letter n Industrial Embedded Systems Resource Guide n PXI, Test & Technology n PXI, Test & Technology E-letter Editorial Director Jerry Gipper [email protected] Contributing Editor Don Dingee Technical Editor Chad Lumsden [email protected] Associate Editor Jennifer Hesse [email protected] European Representative Hermann Strass [email protected] Special Projects Editor Bob Stasonis Senior Designer Joann Toth Senior Web Developer Konrad Witte Graphic Specialist David Diomede OpenSystems Circulation/Office Publishing™ Manager Phyllis Thompson [email protected]

OpenSystems Publishing™OpenSystems Publishing Editorial/Production office: 16872 E. Avenue of the Fountains, Ste 203, Fountain Hills, AZ 85268 Tel: 480-967-5581 n Fax: 480-837-6466 Website: www.opensystems-publishing.com Publishers John Black, Michael Hopper, Wayne Kristoff Vice President Editorial Rosemary Kristoff

Communications Group Editorial Director Joe Pavlat Assistant Managing Editor Anne Fisher Senior Editor (columns) Terri Thorson Technology Editor Curt Schwaderer Associate Editor Jennifer Hesse European Representative Hermann Strass

Military & Aerospace Group Group Editorial Director Chris Ciufo Assistant Editor Sharon Schnakenburg Senior Editor (columns) Terri Thorson European Representative Hermann Strass European Bureau Chief Stefan Baginski

ISSN: Print 1542-6408, Online 1542-6459 Embedded Computing Design is published 8 times a year by OpenSystems Publishing LLC., 30233 Jefferson Ave., St. Clair Shores, MI 48082. Subscriptions are free to persons interested in the design or promotion of embedded computing systems. For others inside the US and Canada, subscriptions are $56/year. For 1st class delivery outside the US and Canada, subscriptions are $80/year (advance payment in US funds required). Canada: Publication agreement number 40048627 Return address: WDS, Station A, PO Box 54, Windsor, ON N9A 615 POSTMASTER: Send address changes to Embedded Computing Design 16872 E. Avenue of the Fountains, Ste 203, Fountain Hills, AZ 85268 RSC# @ www.embedded-computing.com/rsc

/ July 2006 Embedded Computing Design RSC# @ www.embedded-computing.com/rsc Embedded technology: Good for what ails you Jerry Gipper

ne industry that stands the most to gain from embedded n Lack of standards computing is health care. Adoption of technology brings n Government regulation or lack thereof immediate benefit to people of all ages and classes worldwide. No other field comes to mind that can bene- The great news is that the community of embedded computing fit so much from even the smallest of investments in suppliers can address these concerns and help the health care Oembedded computing technology, technology that enables better industry break through these barriers. It is encouraging to see how treatment at lower costs. many embedded computing suppliers have identified health care as a major market focus. Health care has become a major topic of discussion these days, as costs continue to rise and concerned consumers wonder if they While it may be many years before we get to the level of health will be able to afford care. At the same time, technology aids care provided in those Star Trek movies, we can expect some advancements in the level of care offered. Technology can be pretty incredible things in our own lifetimes. found in all aspects of patient care from diagnosis, monitoring, and treatment to patient record keeping. Doctors use embedded I look forward to the day when I can walk into a doctor’s office for computing to take your pulse, monitor your heart rate, analyze the first time, scan my retina (or whatever the biosecurity system your blood, and scan your body. Robotic surgery is becoming of the day is) to log in to the worldwide patient database, not available, giving surgeons heightened precision not possible spend 30 to 60 minutes filling out patient history (most I forget) before and enabling all types of minimally invasive surgeries. and information (most I don’t know) that is duplicated on several forms, and immediately see a doctor. During the appointment, Telerehabilitation and home care both use existing computing the doctor pulls up my latest full-body scan on the life-size technologies, including PCs, wireless networks, even video games; high-definition body scan display and DNA/blood/gas analysis, products that many of us have in our homes today. Body scanning discussing my current ailments and warning me of potential health from new, very low-dosage X-ray, Magnetic Resonance Imaging care concerns. (MRI), and Computed Tomography (CT) imaging modalities draws heavily from embedded computing technology. After the doctor annotates my electronic record, I stop at one of the four drugstores on the corner to pick up my electronically Taking a cue from the IT industry, patient record keeping and submitted prescription. I later head to the gym where I log in to sharing may soon be widely implemented. If we simply automated their system, again with the same technology, and pull up the record keeping in the United States, it is estimated we could save latest exercise program as recommended by my doctor. I stop at more than $236 billion annually, broken down as follows: the grocery store on the way home to pick up some items, per the diet prescribed by my doctor, which I review on the secure kiosk n $77 billion moving from paper to digital patient records at the store or my PDA. n $30 billion moving from paper to electronic billing and claims processing I would enjoy hearing your comments on embedded computing n $29 billion moving from handwritten to electronic drug applications of the future. Tell me about some of the incredible prescriptions ways we might expect embedded computing to change our lives n $100 billion in eliminating insurance fraud with electronic 10-20 years from now. I would be happy to feature some of your records ideas in upcoming issues. Your suggestions and comments are welcome. Please contact me at [email protected]. Though privacy advocates argue that too many people would have access to such an extensive database of personal records, technology and future breakthroughs could minimize or eliminate the risks.

Health care faces many of the same barriers to widespread acceptance that other industries confront, such as: Jerry Gipper Editorial Director n Lack of appropriate and useful software that reflects an understanding of the business processes of hospitals and the organizations that feed into them

/ July 2006 Embedded Computing Design RSC# @ www.embedded-computing.com/rsc RSC# 10 @ www.embedded-computing.com/rsc By Don Dingee All these things that I’ve ...

eople talk about pervasive electronics, but it is staggering to The cashier is baffled, asking if I had bought this item and that step back and look how far we have come. Everything we item, and if I want the free pack of bottled water. I then realize the Ptouch now is electronic in some form. These changes have point-of-sale system has just made an offer based on the helpful changed us, whether we noticed or not, mostly for the better. woman’s profile, not mine. Five minutes, another cashier, and a manager later, they decide it’s easier to give me the discount and skip While there are many electronic entertainment and productivity the promotional item than it is to try and start the transaction over. devices, the changes I’m talking about are found in routine, everyday things. My Sonicare toothbrush has a microprocessor controlling the time it runs and the force it applies. My German Shepherd has a chip that uniquely identifies her with the wave of a scanner. And that’s just the beginning. I’ve wondered if there’s room for one

I’ve got sole more chip. But, I hadn’t imagined years Nike and Apple observed that runners often carry iPods for entertainment, so wouldn’t that be the place to create a runner’s ago that one might show up in my shoe information center? Enter the Nike+iPod Sport Kit: n A wireless sensor insert weighing 0.23 ounces containing an or my 12 pack of Dr. Pepper ... accelerometer package and a 2.4 GHz wireless transmitter operating on a proprietary protocol n A Nike+ shoe with a specially designed pocket beneath the Does all this technology really help me? I hope the presence insole to hold the sensor insert of Radio Frequency Identification (RFID) on everything soon n An Apple iPod Nano with a small wireless receiver attached doesn’t make it even harder to buy stuff. Just take my money, give to its docking port, and the software included in the kit me my Dr. Pepper, please, and keep things simple.

The kit turns the iPod into a sports entertainment and informa- When there’s nowhere else to run tion device. Time, distance, calories burned, and pace data are Getting into a car shows us just how pervasive electronics stored on the iPod and displayed on screen; real-time audible really are. I can’t enter my car easily in the first place without feedback also is provided through headphones. Soon, using the that little device on my key chain. And it won’t start without nikeplus.com personal service website, runners will be able to the key because it contains a chip. The seats are powered and upload, track, and analyze their workouts. remember my preferences. The lights sense conditions and come on automatically, and the mirror dims automatically to cut glare. But, I’m not a shopper The transmission has learned my driving habits and tuned its A recent Associated Press – Ipsos poll identified the checkout line shift pattern. A whole range of other things is going on in the at our friendly local grocery store as the place where our patience power train, braking, safety, entertainment, and navigation sys- is most likely to be tested. I can relate. tems. We’ll be talking about transportation and telematics right here in the November issue of Embedded Computing Design, so Buying a 12 pack of Dr. Pepper the other day, the cashier dutifully stay tuned. asks for my frequent shopper card. My wife is out of town, and she carries the card for this store. The cashier has a backup plan: With all this in mind, I’ve wondered if there’s room for one more What’s my home phone number? Unfortunately, our phone number chip. But, I hadn’t imagined years ago that one might show up isn’t on the record for our card in their system, so I decline. in my shoe or my 12 pack of Dr. Pepper, so it seems there are many more possibilities to run with embedded computing. We’ll The helpful woman in front of me overhears this conversation be covering them as they develop. while she’s multitasking on her Motorola Bluetooth headset and Sprint phone, and instantly offers her phone number for If you haven’t got it already, a good song for your Nike+iPod the cashier. She’s just being helpful, of course, but she clearly is The Killers’ “All These Things That I’ve Done,” and thanks knows the game – buy products, and the store tracks her purchase to them for the lyrical references. Don’t put me on the back patterns, offering discounts and free stuff. When something’s on burner; e-mail your thoughts and ideas to ddingee@opensystems- sale, privacy takes a backseat. publishing.com.

Embedded Computing Design July 2006 / 11 Embedded medical and biological applications By Hermann Strass

Medical care in real time An Italian team of doctors and embedded electronics experts at the Marcella Istituto Superiore di Sanita in Rome has developed a telerehabilitation (remote rehabilitation) system that enables patients with neurological problems to complete their convalescence training, for example, after a stroke, at home under real-time supervision by the experts. Duration and intensity of the exercises as well as physiological data, such as blood pressure or Electrocardiogram (ECG), are monitored at the local hospital. During exercises, such as using biofeedback and video conferencing, or after analysis of a training unit, doctors can recommend further or different training to their remote patients. This saves costs and time, and allows more patients to receive professional care with online technology.

The medical department of Ludwig Maximilian University (LMU) of Munich, Germany, is cooperating with the Health Foundation Hospital in Noerdlingen in testing the exchange of real-time

Figure 1

medical images and other medical data during an operation or other emergency tasks. LMU not only educates medical students, but also operates one of the most highly regarded hospitals in Germany. The hospital in Noerdlingen is small and old, founded in the 13th century, but very modern. It serves several small communities around it.

Noerdlingen is fairly remote by German standards, which is one reason why it was chosen for this application. So if an operation is or becomes critical, experts in Munich can literally see what is going on in the operation room in Noerdlingen and give recommendations on what to do or analyze the available data, including Computed Tomography (CT) scans. A number of embedded electronic devices (blood pressure, heart frequency, ultrasound, laser imaging, nerve reaction speed, and so on), high- performance computer systems, and high-speed communications links are involved in this application.

Abuzz over passive transponders The Free University Berlin and the Rothamsted Research Institute, United Kingdom, are working together, using harmonic radar (see Figure 1) to discover how bees find food. If bees relied on random searching for food, they would starve. With random searching, the odds of finding food are about 3 in 200 million, considering that bees fly about 6 km (3.75 miles) away from their beehive to look for food.

To demonstrate that bees have a method to their madness, experts have developed a completely passive, that is, no batteries needed, transponder weighing 3 mg, or 1.5 percent of an average bee’s RSC# 12 @ www.embedded-computing.com/rsc weight. About 9,300 such devices would equal one ounce. The

12 / July 2006 Embedded Computing Design passive transponder uses the European events energy from an incoming The European Telemetry Conference, also known as radar signal, transposing it the Garmisch Conference, was held at its usual location in into an outgoing signal at Garmisch, Germany, May 2-5. Traditionally, this conference double the incoming fre- focuses on aerospace/avionics, car racing, and similar applications. quency. Sensors can easily However, most presentations this year pertained to biotelemetry. detect such a signal. Since a Presentations on biotelemetry applications ranged from obser- normal reflected signal from vations of birds (worldwide), salmon (in the United States), an individual bee would be or cheetah (from Namibia) migrations to a number of human too weak and undistinguish- telemetry applications, including remote monitoring of patients in able from reflections from dangerous health conditions or patients in telerehabilitation. Other plant leaves, the transponder- applications included long-term monitoring of intracorporeal enhanced signal is easier to (inside the body) implants using wireless inductive powering and track via radar. Researchers other telemetry applications on humans. hope they can use the results Figure 2 to propose theories that ex- For more information, contact Hermann at: plain how bees find their food. [email protected].

3D audio assistance HARTING Mitronics AG, Switzerland, a company of the HARTING Group, Germany, is noted for its expertise in Molded Interconnect Device (MID) technology, a 3D contour-following electromechanical interconnect technology. A hearing aid (Figure 2) currently in production at Siemens Audiological (Medical), Germany, using MID technology was runner-up for the Hermes Award for product innovation at the Hannover Messe Fair, and won the MID Industrial Prize at Productronica in Munich. With MID technology, Siemens Medical can use three directional stereoscopic microphones and ear-to-ear wireless communication to enable simultaneous 3D hearing with optimized adaptive noise cancellation in a device smaller than traditional hearing aids, so it is practically embedded in the human ear. HARTING Mitronics used a laser system from LPKF, Germany, in a laser direct structuring process to produce the MID devices.

If an operation is or becomes critical, experts in Munich can literally what is going on in the operation room in Noerdlingen ... RSC# 13 @ www.embedded-computing.com/rsc

Embedded Computing Design July 2006 / 13 PERSPECTIVE AND NEWS Building embedded software with the Eclipse Process Framework

ystems relying on embedded software typically demand high reliability. Consider the common embedded software domains of Savionics, automotive, and medical devices, where mistakes can cost lives.

As a result, embedded software develop- By Chris Sibbald and Kurt Sand ment processes must provide the checks and balances necessary to ensure high reliability software that satisfies regulatory the EPF Composer user interface (authoring perspective) that process authors would use requirements, such as RTCA DO-178B or in developing process descriptions. FDA 21 CFR part 820. But, development processes cannot hinder innovation, or a The tailored configuration can then be published to HTML for dissemination and exported company will lose its competitive edge. to Microsoft Project to create a project plan. Planned integrations with configuration and change management tools will enable a controlled, auditable evolution of the process How does one balance the need for rigor- content itself. The rest of the development team does not need to know the details of how ous process and the need for unbridled the process was created; they can simply view the end product. Figure 2 shows a sample creativity? The Eclipse Process Framework page from the published website that development team members would use to obtain (EPF) delivers a process framework that guidance in applying the process. supports such a compromise. The exemplary processes, OpenUP (based upon the Unified Process) and Agile component What is EPF? (based upon Agile methods), are designed to be minimal, complete, and additive. They EPF is an open source project within the are minimal in the sense that only fundamental content is included; complete in that they Eclipse Foundation that aims to provide can be manifested as an entire process for the development of software on any platform an extensible framework, which includes for any domain; and additive in that they can be used as the foundation on which process exemplary processes and a tool for process extensions can be easily added to meet the needs of a specific situation. engineering.

The tool, EPF Composer, provides an environment for defining, tailoring, managing, and communicating development processes. Process authors can define who does what, when, and how. They can provide guidance, templates, and other supporting material and publish the information as a website on a corporate intranet.

By separating the method content, which defines the roles, work products, tasks, and associated guidance from the process, which defines the phases, iterations, activities, and tasks in the context of a work breakdown structure, one can reuse content across projects, phases, and iterations, tailoring the life cycle to meet the specific needs of the project team. This approach effectively supports top-down process definition (phases, iterations, and activities), bottom-up process definition (roles, work products, and tasks), or a combination of the two. Figure 1 shows Figure 1

14 / July 2006 Embedded Computing Design variants that address different domains. Potential sources of these out-of-the-box processes include:

n An Eclipse open source community that will evolve and improve upon the exemplary processes n Tool vendors that wish to provide process support for their tools n Academia and research organizations interested in bringing state-of-the-art processes into the mainstream n Standards bodies looking for efficient process delivery mechanisms (ISO 15288, 12207, and others) n Partners, suppliers, and customers with whom we must collaborate

EPF provides an efficient mechanism to communicate processes within the organization and with external stakeholders. Many contracts require a description of internal processes as part of any bid. Standards and regulatory Figure 2 bodies, such as ISO, RTCA, and FDA, require documentation of the process for certification (such as the ISO 9000 Needs of embedded software development series). Noncompliance can have drastic EPF is as applicable to a small software development team working on an embedded consequences. Similar requirements for device as it is for a large aerospace and defense contractor building a system of systems. documenting the development process are applicable in the IT domain as well, All teams need to define roles, responsibilities, work products, and work flows to function Sarbanes-Oxley regulations, for example. cohesively as a team. All teams need to capture corporate knowledge and train new hires. The only difference is the scale of the project (size of team, scope of work) and level of detail (descriptive versus prescriptive) that must be captured and communicated. EPF provides an Smaller teams generally require, and want, less process overhead since communications and coordination are easier in a small team environment, and they simply do not have the resources for any additional overhead. efficient mechanism to

However, the larger the team and the higher the cost of failure, the more important it is communicate processes to have a well-defined process with suitable checks and balances to ensure the efficient and effective development of high-quality products that meet stakeholder and regulatory within the organization requirements.

The ability of the framework to accommodate both extremes, using a standard tool and with external and language, enables the process engineer to dial in the appropriate amount of control while balancing the need to remain agile and creative. For example, mechanisms called stakeholders. variability within EPF Composer permit one to select existing content for reuse, tailor existing content, replace existing content with new content, or remove existing method content from a configuration in order to strike the balance required for any particular Connect the dots: People, process, situation. and technology Organizations depend upon three key Benefits of EPF resources for success: People, process, The availability of an open source framework for process engineering and implementation and technology. Process definitionshould provides a number of significant benefits to organizations. The following discussion precede process automation since invest- provides a limited sample of some of these benefits. ments in technology and tools are wasted if people do not know when and how to Capture and communicate corporate knowledge use them. The EPF ability to integrate tool Organizations cannot depend upon the knowledge and heroics of individuals for their guidance and extended help simplifies the success. The ability to quickly and easily capture and communicate corporate knowledge integration of these three resources. reduces the risk of losing key personnel and provides an effective means of training new team members. Process improvement initiatives such as CMMI are simplified. The current Integrate best practices from multiple sources process can be documented, baselined, By providing a common framework, it is easier to integrate best practices from multiple and analyzed to identify gaps and areas sources. As the EPF ecosystem evolves, we expect to see many out-of-the-box process for improvement. Subsequent process

Embedded Computing Design July 2006 / 15 improvement initiatives are simplified via configuration and change management of the process artifacts.

As mentioned earlier, organizations are often required to state what they do and do what they state in order to satisfy regulatory or certification requirements. The EPF provides the ability to capture, communicate, and evolve this information as a model. Compared to the traditional paper-based process descriptions that typically become obsolete and get ignored, information in the EPF is easily referenced and updated.

Improving development processes The EPF project represents a collaboration of industry and academic thought leaders dedicated to providing an open source framework that supports process engineering and implementation.

The framework provides a standard language and tool for developing processes of any kind, be they lightweight software development processes for small teams working on internal IT projects, detailed prescriptive processes for large distributed teams working on a global C4ISR system of systems, or organizations wishing to optimize business processes.

In addition to the exemplary processes that will form part of the initial and subsequent EPF releases, a number of other domain-specific processes likely will become available from various sources outside of the Eclipse Foundation, creating even more value for those looking to improve their development process.

Visit the EPF website at www.eclipse.org/epf to learn more and get involved.

A final thought: Although the Eclipse platform is the technology behind the EPF, the EPF is not limited to developing processes for Eclipse-based software development. The EPF is applicable to any process, including business processes, technical support processes, as well as systems and software development processes.

Chris Sibbald, PhD, is the senior systems engineer with Telelogic. He has more than 10 years of systems engineering experience, including a former position as payload systems engineer for the Canadian Space Agency’s RADARSAT spacecraft. Chris was also president of Objective SST, an engineering consultancy in Ottawa that provides software and systems development process mentoring and consulting. Chris holds a PhD in Electrical Engineering from the University of Ottawa, and can be reached at [email protected].

Kurt Sand is a senior manager at Telelogic, where he defines, develops, and promotes software development and governance and compliance solutions for Global 1000 companies. Kurt has more than 15 years of experience in the IT industry as a software developer, project manager, and business unit manager. Kurt holds BS and MS degrees in Computer Engineering from Lehigh University and Syracuse University, respectively, and has conducted postgraduate studies in Engineering Management at George Washington University. He can be reached at [email protected].

RSC# 16 @ www.embedded-computing.com/rsc Upcoming E-cast: AUGUST 2006

Upcoming E-casts: RapidIO COM Express OpenSystems DSP/FPGA Reconfigurable Computing Publishing™

Registration and archived E-casts available at: www.opensystems-publishing.com/ecast

16 / July 2006 Embedded Computing Design RSC# 17 @ www.embedded-computing.com/rsc

As cyber crime has escalated to the point

that hackers make millions from security breaches, the need for a new approach to designing tamper-resistant systems has become urgent. Dan explores the reasons why many operating systems today are vulnerable and recommends a microkernel architecture that satisfies the requirements of a secure, survivable system.

Quick. Which of these events really systems inherently tamper resistant and attack on a life-critical embedded control happened: capable of surviving assaults. Otherwise, system can be catastrophic. Consequently, we are simply erecting concrete barriers cyber extortionists consider such systems a) Computer worm crashes safety system around a house of cards. prime targets. in Ohio nuclear plant b) Virus halts train service in 23 states The need for such an approach has been Truth be told, the principles of creating a c) Young recluse cracks computers that made all the more urgent by a major design inherently survivable and tamper control California dams shift in cyber crime. Yesterday, hackers resistant aren’t all that new. In fact, many d) Hacker uses laptop to release 260,000 cracked systems for thrills and notoriety; of them were established as far back as the gallons of raw sewage today, they do it for profit. It has become 1970s, when researchers such as Saltzer & a full-time job, staffed by dedicated Schroeder published seminal papers on the The answer, sad to say, is all of the professionals. If a hacker stands to make topic. The surprise is how much, and how above. These attacks, and thousands money by accessing your data, or by long, the software industry has ignored like them, demonstrate that building a threatening to launch a Denial-of-Service them. This omission helps explain why secure perimeter around our computer (DoS) attack on your system if you don’t our servers and desktops are so vulnerable systems no longer provides enough safety. pay an extortion fee, then you’re a target. to malicious exploits and why many Firewalls, intrusion detection software, embedded systems are equally at risk. and antivirus programs are all important, Worse, these professionals are targeting but no matter how robust a perimeter not only corporate IT servers, but also Consider the key principle of least pri- they create, hackers can and will break control and supervisory systems – systems vilege, which states that a software com- through. Instead, we need a new approach that keep factories running, power flowing, ponent should only have the privileges it to designing the systems we want to and trains from derailing. An attack on a needs to perform a given task, and nothing protect, an approach that can make those corporate server might be costly, but an more. If a component must, say, read data, from the ground up

By Dan Dodge

18 / July 2006 Embedded Computing Design but doesn’t need to modify that data, then it shouldn’t be granted write privileges, either explicitly or implicitly. Otherwise, that component could serve as a leverage point for a malevolent exploit or software bug.

As it turns out, the majority of operating systems today seriously violate this principle. For instance, in a monolithic kernel such as Windows or Linux, device Figure 1 drivers, file systems, and protocol stacks all run in the kernel’s memory address protect critical processes and thwart DoS When it comes to building secure, sur- space at the highest privilege level. attacks. vivable systems, what you start with Each of these services can, in effect, do determines what you end up with. anything it wants. Consequently, a single None of the scenarios mentioned earlier Fortunately, the underlying principles programming error or piece of malicious caused serious harm, with the possible we must embrace aren’t unproven or code in any of these components can (and pungent) exception of the sewage obscure, but simply good, well-accepted compromise the reliability and security incident. However, they demonstrate the programming practices. The groundwork of the entire system. Imagine a building phenomenal trust we place in complex, has already been laid; let the next where a crack in a single brick can bring software-controlled systems, and how generation of innovative – and secure down the entire structure, and you’ve got vulnerable we become if those systems – systems begin. the idea. are compromised. As software designers, developers, and managers, our task, then, Dan Dodge, In response, many embedded system is to create systems that are inherently Cofounder and designers are adopting a more modular trustworthy. CEO, QNX Software OS architecture, where drivers, protocol Systems, with business stacks, and other system services partner Gordon Bell run outside the kernel as user-space created one of the processes. This microkernel approach world’s first message- not only allows developers to enforce passing operating the principle of least privilege on system A system must systems, the fore- services, but can also result in a tamper- runner to the QNX resistant kernel that hackers cannot bend Neutrino microkernel Real-Time or modify. This approach can also satisfy guarantee that existing Operating System (RTOS). Dan also other requirements of a secure, survivable established the QNX-in-Education system, such as fault tolerance, in which software tasks always program, which offers colleges and the system will operate correctly even universities free software for classroom if a driver faults, and rollback, in which work and research projects, and served the system will undo the effects of an have the resources as a founding steward of the Eclipse unwanted operation while preserving its consortium. His honors include the integrity. J.W. Graham medal in Computing they need, even if and Innovation from the University of Microkernel technology with secure Waterloo, an award shared by industry partitioning, which gives applications an untrustworthy luminaries such as Bill Reeves of guaranteed access to computing re- Pixar Studios and Jim Mitchell of sources, should be used in virtually any Sun Microsystems. Dan holds an scenario. The need for such guarantees is application or DoS MS degree in Mathematics. especially urgent in the embedded market. Keeping pace with evolving technologies attack attempts to To learn more, contact Dan at: requires the ability to download and run new software throughout an embedded product’s life cycle, for instance, in-car monopolize the CPU. QNX Software Systems telematics and infotainment systems. In some cases, this new software may be 175 Terence Matthews Crescent untrustworthy, an added risk. To address such concerns, a system must guarantee But, trustworthiness isn’t simply an add- Ottawa, Ontario that existing software tasks always have the on layer. It must be built from the ground resources (for example, CPU cycles) they up. Start with a software architecture Canada, K2M 1W8 need, even if an untrustworthy application that embraces fundamental principles of Tel: 613-591-0931 or DoS attack attempts to monopolize the security, such as separation of privilege, CPU. Properly implemented, resource fail-safe defaults, complete mediation, Fax: 613-591-3579 from the ground up partitioning can enforce those guarantees and economy of mechanism, and you’ve without any need for software recoding or got a major head start. Fail to do so, and E-mail: [email protected] extra hardware. As illustrated in Figure 1, you fight a costly, uphill battle. For proof, partitioning guarantees computing re- consider the endless parade of patches Website: www.qnx.com sources such as CPU time and memory to needed to secure our desktops.

Embedded Computing Design July 2006 / 19 RSC# 20 @ www.embedded-computing.com/rsc Mobile phone security –

By Seiji Inoue

n recent years, mobile phones have become widely used in many countries. However, these phones clearly are not just used for conversations, but also as information terminals for e-mail, Web searches, and a variety of Internet services. They also meet consumers’ entertainment needs for playing games, viewing movies, watching digital TV, and a host of other functions. Currently, mobile handsets are equipped with the same functionality as credit cards and railway commuter passes. On the whole, the The problem with passwords importance of mobile phones in daily life has been rapidly increasing. The mobile phone market is keenly interested in terminal security. Due to the risks of significant loss or damage simply from losing a mobile phone, the market On the other hand, the illegal use of mobile phones has also risen, needs a practical means of dealing with the problem of handset loss. Users forgo and there is a growing risk that e-mail addresses and other personal the password lock function because of information can fall into the wrong hands when handsets are lost or the inconvenience of entering a password each time the phone is used. Most phones stolen. Mobile phones need a reliable and easy-to-use security system produced today only have the password security option and lack fingerprint to be accepted by users. An identification system based on facial rec- recognition capability. One common element that many phones now have, a ognition succeeds in protecting personal information and preventing built-in camera, can be used as part of a facial recognition security system to add its improper use when a mobile handset is lost or stolen. a user-friendly security system to existing

Embedded Computing Design July 2006 / 21 costs, making the software attractive in regards to the ease of decision-making. It is necessary to perform the design work independent of I/O and memory management.

Requirement 3: Processing power Unnecessary code must be omitted since there is never enough processor power. Ultimately, if software runs slowly or not at all, a portion may be put into assembler code, and measures must be considered to make improvements in the cache hit ratio. Having completed this, developers should consider writing the embedded software in C language as much as possible because, even if the software is partly written in assembler, it becomes necessary to perform the development work while marking comparisons with the C language-based code to discover the difference with the code optimized by the compiler. However, since the transformation into assembler depends on the processor, it is preferable for optimization to be possible in C language Figure 1 without any changes. phones without adding devices. Figure 1 Requirement 4: Quality demonstrates how the face recognition Naturally, mobile phone manufacturers perform thorough testing prior to shipment to function works when opening a mobile avoid the discovery of bugs after mobile handsets have been shipped. Nevertheless, bugs phone. can still turn up even after shipment, which is why a company’s quality record is so important in this market. If the program is already in use by a mobile phone manufacturer, Software obstacles an objective determination can be made about the stability of the program and the quality Adding software to a mobile phone to level of the software for the time it has been used. permit using the internal camera for face sensing creates several challenges. Major Market challenges and the solutions required carriers have requirements that programs Effectiveness of the inside camera address these roadblocks. Many new phones now have two cameras. An external camera captures images and videos, while the internal or inside camera (see Figure 2), usually located under the display, Requirement 1: Code size captures images to be sent during a video call. In Japan, NTT DoCoMo’s FOMA phone To succeed in this market, the first has become very popular, as more than 20 million people signed contracts for TV phones challenge is whether or not the software as of 2005 year’s end. However, the TV phone function in a mobile phone is actually not can be loaded into the mobile phone. used that much. The challenge, then, is to find an effective use for the inside camera. This The program must be small enough to camera is provided in TV phones, but if used only for that function, it represents nothing comfortably fit in the space available on more than a cost from the standpoint of the terminal manufacturer. the mobile phone. If developers want to provide usable innovations, they must Despite the fact that outside cameras have advanced to the point that they now have work as hard as possible to reduce the size resolutions of 3 megapixels, inside cameras have hardly made any progress, and still of the software. The first question phone have resolutions of only about 100,000 pixels. Telecom carriers have devised TV phone manufacturers typically ask is, “How big services to increase revenue streams through fees, but in light of the actual situation, is the code?” some mobile phone manufacturers have proposed to carriers designs in which the inside camera has simply been eliminated. In Japan, the carriers determined the inside camera Looking at the world as a whole reveals specification, so the industry must find a good way to use the inside cameragiven the wide range in evolutionary speed this situation. of mobile phones in each country. The processing speeds and program sizes Challenges to users required by embedded software have to be In Japan, the Act for Protection of Personal Information, a law mandating measures to as fast and as small as possible. And when protect against unauthorized disclosure of personal information held by corporations, the customer is a wireless phone carrier, not a mobile phone manufacturer, the software must be suitable for installation on all mobile phones, from high-end to low-end models as provided by that carrier.

Requirement 2: Software portability Licensed software must have the ability to operate without trouble in the operating systems and on processors used by each mobile phone manufacturer. By designing programs with a high degree of portability, program providers can offer software with no compromises in software quality. Likewise, mobile phone manufacturers considering the adoption of a particular program can perform a rapid evaluation without having to spend much on porting Figure 2

22 / July 2006 Embedded Computing Design Using the inside camera for face recognition fulfills the need to protect mobile phone user information. came into force in 2005. Other countries are taking note and following suit with their own legislation. The amount of customer information stored in mobile phones used for business purposes has increased dramatically. In parallel, however, mobile handsets are often left somewhere or simply dropped. Companies typically advise employees to lock their phones with passwords, but in reality, passwords are not used.

Face sensing technology for the mobile handset market Face Sensing Engine (FSE), developed by Oki Electric, is a user authentication solution that uses face recognition technology. FSE is middleware developed to be embedded into mobile terminals to process images of individual faces. Vodafone K.K., for example, has adopted a face recognition function using facial characteristics for its 3G mobile handset, which began shipments this spring.

With the user authentication solution using FSE, entering a password becomes unnecessary, and simply looking at the mobile handset camera performs user authentication, as shown in Figure 1. Once the owner of the handset has been verified, all the internal data in the handset can be accessed, and phone calls can be made as well. But, the handset will be locked if it is determined that the individual attempting to use the phone is not registered as the owner. The handset will be locked, the internal data cannot be accessed, and calls cannot be placed.

This is far more convenient than access control by means of passwords, and puts into users’ hands mobile phones offering a far higher degree of security than today’s phones, in which the password lock mechanism is not actually used. Using the inside camera for face recognition RSC# 23 @ www.embedded-computing.com/rsc fulfills the need to protect mobile phone

Embedded Computing Design July 2006 / 23 user information and creates a viable use for inside cameras of mobile handsets provided by carriers.

FSE is not simply an application program; it is also a collection of image recognition algorithms. Developing these algorithms while maintaining the technical requirements made implementation quite a challenge. The designers had to meet the goal of developing an algorithm that would perform face recognition processing 10 times per second, using a processor with a clock speed of just 100 MHz.

However, one may ask about the verification accuracy of FSE. Mobile phone manufacturers and carriers want to know how it compares with fingerprint recognition. They worry that a photograph can be used to trick the system, and some carriers have therefore declined to adopt the technology. Currently, many researchers are studying facial recognition technology, and they would probably say the same thing. But, users are not even locking their phones by using passwords, so compared to the present state, the use of FSE would represent a major increase in security without changing the way people actually use their mobile handsets. Anyone who, for malicious purposes, wanted to gain access to the contents of a mobile phone belonging to someone else would, in the process, have an incriminating photo of his own face recorded on that phone. Figure 3 shows a screen currently under development that verifies facial images. Figure 3 Implementation of FSE FSE is positioned as a core platform of Inside Camera Solutions. Through this middleware, which provides face recognition functionality, a user verification solution can readily be achieved. FSE also offers a variety of auxiliary functions including facial location detection within the image, detection of characteristic feature points, and point-tracking functions. Thus, it provides a wide range of solution possibilities from security to games through the combined use of the solution library provided in the upper layer of FSE.

For instance, FaceCommunicator generates an animated image reflecting the facial expressions of the user in the form of a real-time 3D animated character that moves as the speaker’s face moves when in TV phone mode. This gives peace of mind when the user wants to have a conversation without showing his or her face or the background. By making good use of the inside camera of the mobile handset, FSE offers a practical solution both for handset manufacturers and users.

As the trend toward using phones for personal information intensifies, phone manufacturers and carriers are discussing the concept of the mobile wallet. Security thus will play an ever-increasing role in whatever applications we choose in the future.

Seiji Inoue is senior manager of the sensing solutions development department at Oki Electric Industry Co., Ltd. Since joining Oki Electric in 1986, Seiji has been involved in the development of computer outputs including Window System (WS) and 3D graphics engine for WS and PCs and image processing. He currently works as the project leader to launch an in-house venture company. Seiji studied Electrical Engineering at Nihon University, College of Science and Technology.

For more information, contact Seiji at:

Oki Electric 785 N. Mary Avenue • Sunnyvale, CA 94085 Tel: 408-737-6479 • Fax: 408-720-1921

RSC# 24 @ www.embedded-computing.com/rsc E-mail: [email protected] • Website: www.oki.com

24 / July 2006 Embedded Computing Design RSC# 25 @ www.embedded-computing.com/rsc Integrating secure, standards-based wireless connectivity for networked medical applications

By Shahin Hatamian

As in many other industry segments, providing seamless Driving forces Networking of disparate medical systems can yield major benefits, interdevice connectivity is becoming an increasingly such as improving staff efficiency, helping control escalating costs important factor for today’s medical applications. of care, and reducing risks of potentially deadly medical errors. According to recent research, an estimated 195,000 people die in The medical sector poses some particular concerns in hospitals per year due to medical mistakes, such as administration of incorrect drugs or omission of critical information from patient terms of privacy and security, but also must leverage records. Many of these errors can be eliminated through the use proven standards-based technologies ubiquitously of electronic medical records, a concept the federal government is promoting to ensure timely collection and immediate availability deployed throughout mainstream networking appli- of accurate patient information at every point of service. cations. Shahin describes the overarching trends and In addition to enhancing overall quality of care and improving challenges in medical networking and discusses key operational efficiency, the consistent collection of data right at the patient’s bedside also benefits the institution by capturing every issues to consider when designing embedded service in order to maximize insurance reimbursement rates. Manual record keeping can too easily miss short episodes when a connectivity into medical applications. specialized piece of equipment is brought into the patient’s room,

26 / July 2006 Embedded Computing Design but if the machine automatically collects and uploads all point-of- service data over a Wireless Local Area Network (WLAN), even the briefest usage can be captured into the patient’s record. Networking of medical equipment Patients can also benefit from wireless connectivity. Enhanced data collection during medical visits improves the accuracy of medical records, thus improving care. Billing information is more can best be achieved alongside accurately captured, reducing billing errors. other tangibly beneficial applications Driven by mandates such as the Health Insurance Portability and Accountability Act (HIPAA), safeguarding privacy of patient information is another critical requirement for networked medical within an overall network strategy. applications. Here again, electronic medical records can help achieve the privacy goals by eliminating the need for having paper- based copies at the patient bedside and other locations. The key offerings, and the ROI equation becomes much more attractive. is tightly controlling access at the point of usage and securing the Some hospitals are justifying networks by giving patients Internet data when transmitted on the WLAN to the central repository. access in their rooms for entertainment and communications.

Challenges However, deployment of multiuse hospital networks raises Medical networking faces many of the same issues as mainstream additional challenges, such as maintaining security in a shared deployments, plus a few additional pressures. Budgets for new IT infrastructure and assuring required Quality of Service (QoS) investment are always tight, so the projected Return on Investment levels for high-priority applications. To address these issues, the (ROI) must be tangible and quantifiable. In addition, the need to ability to manage separate, secure Virtual LANs (VLANs) within overcome technology averseness on the part of many doctors and the common physical network is a critical element that must be nurses, as well as the hurdles associated with a highly regulated designed and implemented at the fundamental connectivity point environment, can make it difficult to institute radical changes on a for every piece of equipment on the network. wholesale basis. Therefore, networking of medical equipment can best be achieved alongside other tangibly beneficial applications Implementation issues within an overall network strategy. For instance, a WLAN network For OEM manufacturers of medical systems, a key to success- for medical equipment alone might not be justifiable, but throw ful network integration is the use of embedded connectivity in mobile communications using VoIP to replace pagers, support solutions that support all major wireless networking standards for back office functions, and possibly even patient access service (802.11a, 802.11b, and 802.11g) and wired Ethernet (100BASE-T,

RSC# 27 @ www.embedded-computing.com/rsc

Embedded Computing Design July 2006 / 27 used to automatically trigger alerts or to initiate preset actions. This not only improves local responsiveness at the patient bedside, but also helps reduce network traffic levels and improve overall performance by handling more decision-making at the edge of the network.

The bottom line is that network connectivity in medical systems is destined to become a ubiquitous requirement, just as it already has Figure 1 within most other computing environments. For medical system OEMs, the key to success will involve selecting embedded, 10BASE-T), while also providing built-in security and pro- modular hardware and software solutions that deliver proven grammability for integrating specialized functions. Figure 1 standards-based networking and security functions, while also shows an embedded 802.11a/b/g module that allows medical offering the flexibility to meet application-specific requirements device OEMs to implement secure wireless connectivity with inherent to the medical sector. little effort. Shahin Hatamian is the director of product The availability of all WLAN standards in the same device gives management and engineering at silex OEMs and their customers the flexibility to have both 802.11g technology america. He has more than and 802.11a in the same environment. Leveraging an 802.11 15 years of wireless industry experience overlay network at 5 GHz allows efficient segregation of high- primarily in product management and product priority traffic away from the increasingly crowded spectrum at development. He holds a BSEE and MSEE 2.4 GHz. The embedded wireless solution should also include a from Rutgers University and an MBA from the full set of secure authentication and encryption standards such as University of California, Irvine. IEEE 802.11i. To learn more, contact Shahin at: Programmability and configurability of the connectivity modules represent other important factors for OEM success. For example, module-level support for Java Virtual Machine (JVM) allows silex technology america an OEM to implement custom applications, such as device 15661 Red Hill Avenue, Suite 120 • Tustin, CA 92780 monitoring, data filtering, and so on, that can run independently of the higher-level system functions. Used in conjunction with E-mail: [email protected] multiple General Purpose I/O (GPIO) ports that can monitor Website: www.silexamerica.com internal or external sensors, this local programmability can be

28 / July 2006 Embedded Computing Design RSC# 28 @ www.embedded-computing.com/rsc RSC# 29 @ www.embedded-computing.com/rsc

Model-driven development for medical devices

By Irv Badr

oday, more than ever, the list of features in medical devices continues to rise due to the availability of Tlower-cost components typically integrated onto even the most simplest of medical devices, such as diagnostic and monitoring systems. However, added components render software applications equally complex, placing added burden on the premarket certification activities (510K submission) and compliance in general.

Using a model-based verification and validation process addresses two major issues in developing medical devices. First, the added complexity of the modeled application is easier to manage due to its abstraction as a model. Secondarily, the verification and validation of the evolving system can be accomplished through an executable model and much earlier in the development process, reducing the total development time and shortening the FDA submission process.

The challenge n Compliance with Current Good Besides the typical work flow used for developing embedded software products, such Manufacturing Practices (CGMP) and as requirements gathering, analysis, system design, detailed design, testing, and pro- Good Design Practices (GDP), both ject management, there is one additional challenge in the medical device industry also dictated by the FDA – compliance. Products developed for the U.S. market are regulated by the Food and Drug Administration (FDA) through Quality Systems Regulations (QSR) 21 CFR Traditional approaches Part §820.30, which essentially require[1]: In order to meet compliance goals, medical device development teams start a DHF at n Proper documentation to be maintained in a Design History File (DHF) the inception of a project. The contents of – 21 CFR Part 11 also governs the usage of electronic signatures in DHF to facilitate this file may include informal, handwritten electronic documentation in lieu of paper version requirements and design notes, as well as n International markets to be ISO 13485:2003 compliant and meet the EU MDD 93/42 formal architecture and design documents regulations in the form of printouts and source code

30 / July 2006 Embedded Computing Design excerpts. For many development teams set of deliverables mentioned in the design that are source code centric, the DHF may output. Therefore, design output may in- include requirements documents, ad hoc clude, in the case of system software, a formal design documents, and a large listing of the source code belonging to number of source code listings, reflecting the application. The goal is to disclose the how the expressed requirements are being intended use of the device and ensure that implemented faithfully as source code[2]. specifications have been implemented and QSR requires this work flow, mandat- design goals met. ing that traceability must be established Model-driven between requirements and their exact im- Verifying the design plementation to prove that the device Design input and output milestones are is being used for its intended purpose; integral parts of a medical device develop- however, the method of establishing ment process and are applied to a large traceability is not spelled out. Source code number of devices; therefore, they do is typically submitted since it is readily not specify what methods should be used development available, and most teams also use source for the verification shown in the process code as the medium for representing outlined in Figure 1. Device makers use system architecture and design. This is an array of tools and processes for this pretty typical in the absence of a formal purpose, but most rely on textual re- modeling approach integrated into the quirements documents and source code for medical development process[3]. listings as described for design input and output. Figure 1 includes a design Design input and output verification step where the design output Figure 1 shows a recommended develop- is verified against the design input. ment process from the FDA Design devices Control Guide[4]. Note the fact that QSR Validating the system compliance can be achieved in conjunction Since most teams consider source code with GDP by adhering to the iterative as the ultimate measure of a system’s development steps shown. Much of sys- implementation, running the completed tem verification is performed against the source code on the actual device frequently requirements by measuring design output demonstrates the intended purpose of the against design input[5]. Design inputs are device. This is a source-centric approach, a set of specifications derived initially as validation occurs by executing the from the user requirements with the goal application code compiled for the real that intended use of the device should device. be spelled out as a set of requirements. Design output, in turn, is a set of pro- While effective, traditional methods of cedures defined by the device maker validating the system prove to be cost- ensuring that the completed work tallies ly and error prone. The task of unit with the corresponding design input. This testing the evolving system on the target essentially mandates a need for traceability device may be cumbersome and slow. It between these two milestones. may not be possible to put the machine through all the intended usage scenarios Traditionally, development processes in- without incurring a heavy expense or clude as design input a formal or semiformal logistical obstacles, such as incomplete set of word processor documents and a few or nonexistent hardware, which may ham- models reflecting a set of specifications per the final testing on the device. An against which the system is to be built. This incomplete platform may render wrong input is transformed into the predisclosed results for the tested application.

Design User needs reviews

Design input

Design process

Verification Design output

Medical Validation device

Figure 1 RSC# 31 @ www.embedded-computing.com/rsc

Embedded Computing Design July 2006 / 31

VIM Processor Connectors

C6x As software and system bugs are found while architecture and design activities during unit testing and are corrected, are traced to system verification. All the their associated design and requirements steps shown are performed automatically may need updating to reflect the changed through the requirements traceability and source code. Depending on the size of the executable model features provided by the application, this may become time the modeling tools[1]. consuming and susceptible to errors as some changes may be missed, causing a An example of automated mismatch between the DHF and the actual validation and verification implemented system. In a model-driven process, both architecture and design of the application are carried A new approach using out in the modeling tool[3]. Additionally, automated tools design tools can automatically generate Design controls and QSR don’t require unit test cases from design models, as the actual operational medical device shown in Figure 3. Here, models for both as the basis for system verification and the design and the test case are executable. validation. As far as the FDA is concerned, The first goal is to verify the underlying it is sufficient to simply demonstrate the design, then to test the design, depicting collected data as proof of intended use of the actual usage scenarios for the device. the device. This data can be collected from Most defects and design oversights are the actual device as well as a simulated caught during this model verification execution of the device, where automat- phase. Unit testing can put the design ed application development tools offer through any possible scenario, many of unparalleled efficiency gains. Figure 2 which may be difficult to create on the presents a process centered on validation actual device. In the forthcoming example, and verification. Like the FDA process tests can show the system’s behavior in the described earlier, this process hinges on unlikely event that the patient exhibits no iterative development of design output detectable pulse in the middle of moni- based on design input. Figure 2 also toring his or her blood oxygen levels. traces requirements to system validation, Creating this test case on the real device

Risk management (DHA and FMEA)

Quality assurance and user testing

Figure 2

Risk management (DHA and FMEA)

RSC# 32 @ www.embedded-computing.com/rsc

Quality assurance and user testing

Figure 3

32 / July 2006 Embedded Computing Design Integer Status; << TestComponent >> << SUT >> Integer HR; tc:ControllerTestComponent1 sut:Controller Integer Sp02; Integer csc; idle Sp02Data (10, HR, Sp02)

DisplayPatientData (HR, Sp02)

Sp02Data (Status, HR, Sp02)

arbiter assertTrue (HR > = 40 && HR < = 200) Status arbiter assertTrue (Sp02 > = 55 && Sp02 < = 100)

<0 else

511 else

Display Sp02 Invalid data

127 else Unit testing can put the design

Interfaces::ToPatient::DisplayPatientData (None, None) through any possible scenario, idle

Interfaces::ToPatient::DisplayPatientData (HR, Sp02) many of which may be difficult to

idle

Display the HR and 02 levels create on the actual device. Figure 4 will require stopping a subject’s heart, to Finally, for system validation, test cases are traced to the system’s operational re- which they may object. quirements. This is automated through requirements traceability feature between a formal requirements management tool and the modeling environment. In other words, a fully Defects and oversights detected are im- automated validation and verification process can be established. mediately corrected in the design, thus eliminating the need for manually changing Development tools the code and the design documenta- Telelogic provides life-cycle development tools (shown in Figure 3) that cover the whole tion. This contrasts traditional develop- arena of system validation and verification as required by QSR. DOORS requirement ment, in which defects are fixed inside management tool is used as the basis for creating the design input, and both system the source code after executing it on the requirements and validation tests are managed inside DOORS as structured and traceable real device. This happens later in the sets of objects. development process, and system debug- ging therefore is much slower than a Two different tools provide a modeling solution depending on the nature of the medical model-driven approach. device being developed. For conventional, standalone medical devices that typically use a single board embedded computer and a Real-Time Operating System (RTOS), The example in Figure 4 illustrates a Telelogic’s Rhapsody provides system architecture and design support as well as design for a blood oxygen monitor used automatic source code generation. Rhapsody is capable of supporting out-of-the-box, to measure blood oxygen levels as well major RTOS platforms. It also provides the added capability of target-hosted cosimulation, as pulse rate through a finger clip sensor. which proves valuable during the validation and verification process when target-level The design shows a state machine respon- verification becomes necessary. sible for de-packetizing the sensor data containing blood oxygen levels (SpO2) Telelogic TAU can be used for complex medical devices using multiple platforms, or and pulse rate. By injecting either real or an arbitrary combination of embedded and desktop systems. This may include devices simulated sensor data, the state machine such as Computed Tomography (CT) scanners or other systems that carry an array of shown in the diagram can quickly verify interrelated platforms, including headless desktop computers and monitoring work- correct operation. Furthermore, the test stations running conventional operating systems such as Linux, UNIX, or Windows. case in Figure 4 verifies that both pulse TAU supports language and operating system independent modeling, and can deploy the rate and oxygen levels are within a safe same models on any arbitrary platform using a mix of programming languages. This is range. This test case is run alongside described as Platform Independent Modeling (PIM), where a single set of models can design verifications to ensure patient safety be used on many different or undefined platforms. The concept of PIM, therefore, adds during unforeseen or complex events that another dimension to increased productivity, allowing designed components to be used are difficult to create. in future generations of unknown platforms.

Additionally, an automated requirements The resulting documentation from the automated development process provides an management tool can be used so that integrated paper trail originating both from requirements management and modeling traceability can be established between activities. This paper trail is automatically created by Telelogic’s DocExpress, which has the design components, such as the state visibility into all Telelogic tools used in this example. DocExpress automatically creates machine pictured in Figure 4, and the word processor pages incorporating text and diagrams from the tools used, and is fully requirements. configurable by the user.

Embedded Computing Design July 2006 / 33 A better, cheaper development process When designing medical devices, design guidelines and regulations as dictated by the FDA QSR can be addressed concurrently with best practices in system and software development. This not only decreases development costs, but also facilitates the validation and verification process dictated by QSR, resulting in a more reliable medical device with less chance of failure in the field. Additionally, this provides live content for the DHF, which is automatically managed and produced. The set of life-cycle management solutions from Telelogic aims to automate the development process through the use of requirements management, system and software modeling, and automated, model- based testing tools, including DOORS, Rhapsody, TAU, and DocExpress.

References [1] Badr, I., “Developing Platform Independent Embedded Applications,” Embedded Systems Magazine, July 2005 [2] American National Standards Institute, Cardiac monitors, heart rate meters and alarms, ANSI/AAMI, EC13:2002 [3] Badr, I., “Rapid Development through Agile Modeling,” Telelogic white paper, February 2005 [4] U.S. Food and Drug Administration, Design Control Guidance For Medical Device Manufacturers, March 11, 1997 [5] U.S. Food and Drug Administration, The Quality System Regulation, January 1, 1997, www.fda.gov/cdrh/qsr/01qsreg.html

Irv Badr, currently a senior product marketing manager for Telelogic, has nearly 20 years of development experience in embedded and modeling industries, designing communication infrastructure for medical devices, networking nodes, digital cable transmission, and industrial controls. He also served as technical lead for sales and marketing of RTOSs, and for modeling tools with TimeSys Corp. and Artisan Software. Irv has a BS in Biomedical Engineering from the University of Illinois and a technical MBA from Northwestern University.

To learn more, contact Irv at:

Telelogic 9401 Jeronimo Road Irvine, CA 92618 Tel: 949-830-8022 • Fax: 949-830-8023 E-mail: [email protected]

RSC# 34 @ www.embedded-computing.com/rsc Website: www.telelogic.com

34 / July 2006 Embedded Computing Design RSC# 35 @ www.embedded-computing.com/rsc RSC# 36 @ www.embedded-computing.com/rsc www.embedded-computing.com/products

LEDs Diagnostic support for custom eliminate boards using PCI Express

leakage PCI Express is the de facto parallel bus of choice for many embedded current designs. Custom I/O boards paired up with commercially available processor boards represent a large number of products. Developers of LED lamps will soon custom boards already have enough to worry about without letting the replace traditional bus interconnect also become a point of design contention. incandescent lamps. Burned-out incandescent bulbs can Kozio’s kDiagnostics products support testing the next generation of create safety concerns, require costly maintenance and repair, and PCI Express peripherals, providing engineers with a tool to validate result in significant downtime. Several potential sources of leakage entire designs at the earliest prototype stages. Kozio supports multiple current in a typical circuit, including transient voltage from a relay PCI Express links and provides testing of all memories, buses, and or discharge from a capacitor, prevent LEDs from being used now. connected devices, including basic configuration testing, link protocol Because LEDs are so efficient, a current as low as 0.25 mA can cause testing, and transaction protocol testing, enabling a one-of-a-kind an LED to illuminate. diagnostics solution for product developers in as little as two weeks.

Data Display Products has developed a series of lamp-based LEDs that feature built-in shunt resistors to reduce unwanted illumination caused Kozio, Inc. by leakage current. Well-suited for industrial environments, power www.kozio.com generation, transmission, and distribution applications, the LS Series kDiagnostics is designed to be a direct replacement for incandescent bulbs. RSC# 29962 LS Series lamps offer upwards of 100,000-hour life, shock and vibration resistance, low heat, and low power consumption.

Data Display Products www.datadisplay.com 32-bit LED Lamps RSC# 30711 microcontroller replaces legacy 16-bit QUICC custom microcode microcontrollers

Today, using FPGAs or ASICs provides the easiest way to add Imagine getting 32-bit custom features to a microprocessor. Developers design the microcontroller performance at the price of an 8- or 16-bit functions needed, then program the features into these additional microcontroller. Now picture getting this same microcontroller parts. Wouldn’t it be great if the microcode within the processor within similar power budgets, one of the largest barriers to entry could be modified to create application-targeted microcode for 32-bit processors into 16-bit applications. Deeply embedded solutions for specific customer needs? applications could have compute processing power only found in previously more expensive and power-hungry microcontrollers. Freescale Semiconductor has introduced the Open QUICC Engine developer program for third-party developers and customers that Luminary Micro, a fabless semiconductor company that want to optimize the communications functionality of applications designs, markets, and sells ARM Cortex-M3 processor-based that leverage Freescale’s QUICC Engine technology. The program microcontrollers, is the first to bring embedded developers ARM for allows developers to customize Freescale microcode that runs $1. Their Stellaris family of 32-bit microcontrollers (MCUs) offers on QUICC Engine technology to support industry-standard features such as an analog-to-digital converter and a sophisticated communications interfaces and protocols without having to add motion control unit, as well as large on-chip memories. The entire ASIC or FPGA devices to their applications. Customized microcode Stellaris line of MCUs brings high-performance 32-bit computing developed through the Open QUICC Engine program will be portable to embedded microcontroller applications at a cost equivalent to and scalable across QUICC Engine technology-based processor legacy 8- and 16-bit devices. All of the Stellaris MCUs are targeted platforms and applications. at embedded and industrial applications, such as building and home automation, factory automation, motor control, and industrial Freescale Semiconductor power control devices. www.freescale.com/quiccengine Open QUICC Engine Luminary Micro RSC# 30710 www.LuminaryMicro.com Stellaris MCUs RSC# 30712

Editor’s Choice Products are drawn from OSP’s product database and press releases. Vendors may add their new products to our website at www.opensystems-publishing.com/vendors/submissions/np/ and submit press releases at www.opensystems-publishing.com/news/submit. OSP reserves the right to publish products based on editors’ discretion alone, and does not guarantee publication of any product entries. Embedded Computing Design July 2006 / 37 RSC# 38 @ www.embedded-computing.com/rsc Multi-CPU architectures with Advanced Switching extensions to PCI Express

By Stephen Christo

he latest chipsets and I/O devices are with each other. However, all nontransparent bridging implementations were proprietary in nature and required replacing their PCI buses and embracing one master root complex controlling the system address space and any I/O complexes. Scaling beyond two CPUs PCI Express (PCIe) as the dominant chip-to- (one active, one hot standby) was problematic at best, chip interconnect. As with PCI, developers are and the reliance on a master root complex meant that the T systems were never truly fault tolerant. working to take advantage of this low-cost, ubiqui- Express In contrast, ASI was developed to connect multiple tous interconnect in multiple embedded applications. PCIe-based root complexes and I/O devices in the The Advanced Switching Interconnect (ASI) is an same topology while maintaining backwards hard- PCI ware and software compatibility to PCIe. As shown in extension of PCIe that allows CPU-to-CPU commu- Figure 1, multiple virtual servers can be formed, each one containing a host CPU that controls one or more nication and dynamic I/O configurations to operate I/O devices without any changes to device drivers, within a PCIe framework. By unifying the CPU-to- CPU and CPU-to-I/O transport structures with ASI, dramatic improvements in performance, system cost, flexibility, and high availability can be realized.

Multi-CPU systems with PCIe and ASI Like PCI, PCIe always has a single root complex that handles all interrupts, maintains responsibility for reset distribution, and provides the final upstream port in the PCIe hierarchical structure. Attached to this single root complex can be one or more I/O complexes. However, most embedded systems require two or more CPUs, so how does PCIe translate to this architecture? To do so, a method for isolating independent address spaces while maintaining a way of communicating between subsystems is required.

In the past, this had been solved with nontransparent bridging, which allowed two intelligent subsystems with independent address spaces to coexist. The capability of address translation provided a window for each of the subsystems’ CPUs to communicate Figure 1

Embedded Computing Design July 2006 / 39 BIOS, and so on. Each virtual server has its own address map, independent of the other processing domains in the topology. The slot dependencies and hierarchical restrictions that one en- ASI’s ability to tunnel PCIe counters with a standard PCIe architecture are no longer valid when tunneling PCIe packets through an ASI fabric. In addition, traffic greatly enhances the the ability to dynamically perform CPU-to-I/O mappings enables a level of system flexibility and availability that could not be considered using only PCIe. flexibility of a platform’s

ASI also extends the interprocessor communication capabilities CPU-to-I/O communication. of PCIe. Nontransparent bridging solutions had proprietary implementations of doorbells, scratchpads, and address translation If a server must be upgraded or I/O requirements change over to perform rudimentary interprocessor communication opera- time, the server-to-I/O connections can be reconfigured through tions. ASI provides several standard methods of CPU-to-CPU system software without the need for system hardware changes, communication including load/store and messaging protocols. such as replugging cards or cables. This software reconfiguration With this capability, the host CPUs of each domain in Figure 1 capability also allows fast failover to redundant servers and/or can concurrently communicate with other host CPUs through I/O modules. Due to this capability, the 2N redundancy model legacy IP/socket or load/store memory-to-memory applications. of the 4U servers is not required. Instead, a lower-cost N+1 redundancy paradigm can be supported; therefore, the number of ASI lowers cost and increases flexibility of HBAs and NICs (four in this example) required in the system is rack-mounted solutions only one more than the total number of servers. The cost of the Although deployments of bladed systems continue to grow, ASI/PCIe-enabled approach is approximately half that of the most server, storage, and embedded solutions continue to use existing solution, and provides a higher degree of flexibility, rack-mounted units interconnected through cables. Processing scalability, and availability. capability per unit area, power consumption, price, and high availability are some of the key factors used in the purchasing ASI’s ability to tunnel PCIe traffic greatly enhances the flexibili- decision. 1U rack-mounted servers are the most efficient in terms ty of a platform’s CPU-to-I/O communication. In addition, ASI of processing capability per unit area, power consumption, and offers other features that can augment system performance. By price, but are typically limited to two I/O expansion slots. With providing orthogonal data movement models, known as protocol only two I/O slots, it becomes difficult to include multiple I/O interfaces in ASI, which are tailored for a specific communication protocols on a processing node and maintain high availability policy, ASI can cost effectively support both CPU-to-I/O device characteristics. By enabling I/O disaggregation, ASI allows solu- communication and CPU-to-CPU interprocessor communication. tion providers to take advantage of the benefits of 1U servers and In the application example of Figure 2, ASI can be extended using continue to offer I/O flexibility and availability. protocol interfaces such as simple load store or simple queuing protocol to enable low-latency, low-overhead, peer-to-peer host Figure 2 demonstrates the advantage of an ASI/PCIe rack- transactions. This allows processing power to be shared across mounted approach. Solutions deployed today could consist of multiple hosts, creating a clustered server environment. three 4U servers (two active and one hot standby) each with redundant Fibre Channel HBAs and Ethernet NICs. The total ASI enables I/O disaggregation in the data center, reducing cost system would require 12U of rack space, six Fibre Channel HBAs, and increasing the processing power per unit area. By introduc- and six Ethernet NICs. In contrast, an ASI/PCIe-enabled system ing peer-to-peer communication, computational power can be consisting of three 1U servers and a single 3U PCIe expansion increased to meet application requirements. Another capability of system would use only half of the rack space. With ASI’s dynamic ASI is the enablement of virtualized I/O where multiple hosts can CPU-to-I/O reconfiguration capability, a given server can connect share a single I/O device. This is true I/O sharing where multiple to one or more I/O devices. hosts can utilize the bandwidth of a single 10 Gigabit Ethernet (GbE) or 8 Gigabit Fibre Channel I/O card. By adding this capability to the CPU-to-I/O relationship through PCIe tunneling, the cost of scaling I/O to a group of servers is significantly reduced. ASI expands opportunity for PCIe in bladed systems Bladed system design has been the norm in communications systems for the past 20 years. With the introduction of IBM’s BladeCenter, HP’s BladeSystem, and Sun’s Netra Blades in the past few years, bladed architectures have migrated to compute applications. To provide such attributes as modularity, flexibility, Current deployment with 4U servers Future deployment with I/O PCIe/ASI disaggregation and fast time to market, standards- based bladed architectures recently have been developed. PICMG in- Figure 2 troduced two bladed architectures

40 / July 2006 Embedded Computing Design for compute and communications applications. The Advanced system with AdvancedMCs providing the processing and I/O Telecom Computing Architecture (AdvancedTCA) standard functionality. Without ASI extensions, the system would use PCIe addresses the carrier-class, high-performance requirements and GbE locally on the AdvancedTCA carrier and only extend of telecom equipment manufacturers. In combination with the GbE to the backplane. Extending this architecture by adding Advanced Mezzanine Card (AdvancedMC) standard, a new level ASI to the backplane provides several advantages, including of modularity and flexibility in I/O, data processing, and memory higher performance, increased flexibility, lower cost, and higher options can be realized. A second backplane standard, MicroTCA, availability. was developed more recently to leverage the AdvancedMC ecosystem and target low- and mid-range telecom/datacom ap- Without ASI, PCIe and GbE are used to connect the AdvancedMCs plications where AdvancedTCA was deemed too costly. locally on the AdvancedTCA carrier card. Due to the constraints of PCIe, only one processor can reside on the carrier card control- IP Multimedia Subsystem (IMS) network elements represent ling the I/O AdvancedMCs. The processor could be located on one of the application areas being addressed by AdvancedTCA, either the AdvancedTCA carrier orExpres on a processors AdvancedMC AdvancedMC, and MicroTCA. Depending upon the element in an in one of the expansion slots. Each solution has its pros and cons. IMS infrastructure (that is, security gateway, application server, By implementing a CPU on the AdvancedTCA planar, all four media gateway, call session controller, and so on), the type and AdvancedMC slots are availablePCI to support I/O. In many systems, mix of processing and I/O elements will change. An architecture such as Digital Subscriber Line Access Multiplexers (DSLAMs) should be flexible enough to support any combination of CPU and Ethernet switches, an important metric is cost per port. By and I/O requirements using the same cards. maximizing the number of I/O, this architecture allows for the lowest cost per port. Disadvantages of this architecture include In general, the following components are required in an IMS the system’s limitation to one processor type, the fact that failure platform: of the CPU eliminates I/O from the system, and lack of flexibility and scalability for those systems that are more processor intensive. n Ethernet for IP-based network infrastructure By moving the CPU to an AdvancedMC, flexibility and upgrade- n Fibre Channel for content storage ability of the processor type can be achieved, but at the expense n Processing (CPU/NPU/DSP) of a 25 percent reduction in I/O density at the system level. n TDM and ATM legacy interfaces By extending the PCIe/GbE architecture with ASI, a developer When looking at the growing ecosystem of AdvancedTCA and can take advantage of the low cost and large ecosystem of PCIe AdvancedMC products, I/O interfaces such as Fibre Channel, GbE, TDM, and ATM are typically provided on AdvancedMCs. Most of the available cards are AMC.1 compliant, routing both PCIe and Ethernet to the AdvancedMC connector. In some cases, like a Fibre Channel card, PCIe is the primary data transport mechanism, and Ethernet, if implemented, is only used for control. In others, like an iTDM AdvancedMC module, GbE is the primary data transport, and PCIe is used for control only.

When developing an IMS platform, most developers want to take advantage of these off-the-shelf modules, but integrating them into a system becomes problematic. Figure 3 shows an architecture for implementing a 5-slot, full mesh AdvancedTCA-based

Figure 3 RSC# 41 @ www.embedded-computing.com/rsc

Embedded Computing Design July 2006 / 41 solutions while expanding system flexibility, modularity, and Taking advantage of the PCIe ecosystem high availability. As demonstrated in the rack-mounted exam- Extending PCIe with Advanced Switching enables high- ple, ASI supports multiple root complexes in a single fabric and performance multi-CPU solutions with reconfigurable I/O has the ability to dynamically reconfigure CPU-to-I/O connec- mappings for compute, storage, and embedded markets. Its tions. With the literal click of a button, a binding can be established compatibility at the physical and data-link layers and ability to between a PCIe-enabled host CPU and one or more PCIe I/O transparently tunnel PCIe allows architects to take advantage of devices. Once the bindings are established, the PCIe domain the large PCIe ecosystem while maintaining a migration strategy created within an ASI fabric can run legacy PCIe applications for scaling to more complex system designs using native ASI without any changes to BIOS, drivers, and so on. Using this features. capability, developers can disaggregate their I/O elements from their processing elements. Stephen Christo is director of strategic marketing for StarGen and chair of the In the AdvancedTCA architecture shown in Figure 3, this capa- ASI SIG Marketing Work Group. He has bility eliminates the disadvantages of the PCIe/GbE solution. more than 15 years of experienceExpres in the s A single architecture can now support systems ranging from electronics industry having held marketing, one CPU and 19 I/O AdvancedMCs to 20 CPU AdvancedMCs. engineering, and managementPCI positions at Whether the application requires CPU-to-I/O or CPU-to-CPU Intel, Digital, CP Clare, and Raytheon. He communication, legacy software is preserved. With ASI and has an MBA from Bentley College and an GbE across the backplane, two interconnects are available for MSEE from Northwestern University. CPU-to-CPU communication. If the GbE pipe is large enough for data traffic between processors, it can be used primarily for For more information, contact Stephen at: that purpose. If higher bandwidths (order of magnitude more) and lower latencies (order of magnitude less) are required, the StarGen ASI fabric can be utilized for CPU-to-CPU data transfers. By 225 Cedar Hill Street, Suite 22 using standard APIs like Internet Protocol over ASI (IPoASI) or a remote RAM disk, legacy software can be preserved for Marlborough, MA 01752 interprocessor communication. Tel: 508-786-9950 • Fax: 508-786-9785 www.StarGen.com Adding ASI to the architecture also increases high availability.

If a processor AdvancedMC fails or needs a software upgrade, the I/O AdvancedMCs under its control can be migrated to Advanced Switching Interconnect SIG another processor AdvancedMC in the system during down time. c/o VTM Advanced Switching does not have the hierarchical restrictions Attn: Nicole Gray of PCIe, and therefore supports the mesh topology of Figure 3 or any other AdvancedTCA topology. In a dual star architecture, 3855 S.W. 153rd Drive • Beaverton, OR 97006 the initial PCIe bindings can be established through the primary Tel: 503-619-0566 • Fax: 503-644-6708 fabric. If this hub card fails or must be replaced, PCIe bindings Website: www.asi-sig.org could be re-established quickly through the secondary hub card.

Solutions for PCI Express design verification By Stephen D’Onofrio, Dr. Ambar Sarkar, Jim Tai, Paradigm Works, and David Lin, Denali Software www.embedded-computing.com/articles/id/?249

Leveraging legacy interconnect with fabric technology By Tom Cox, RapidIO Trade Association www.embedded-computing.com/articles/id/?239

PCI Express emerging as the interconnect standard By John Gudmundson, PLX Technology www.embedded-computing.com/articles/id/?234

42 / July 2006 Embedded Computing Design Express

PCI

RSC# 43 @ www.embedded-computing.com/rsc www.embedded-computing.com/products

By Chad Lumsden

Chips & Cores: ARM 8 MB flash memory • GNU C tool chain • Palm- adjustable gain and filtering, eliminating the size metal box for wall mount and DIN RAIL need for auxiliary signal conditioning equip- Artila Electronics. Co., Ltd. • Direct 9 to 48 Vdc input • ment • Based on the Analog Devices AD9260 Website: www.artila.com Sigma Delta ADC, the ICS-1745 simultaneously Model: Matrix 500 RSC No: 30340 Data acquisition samples up to eight differential input signals at ARM9-based industrial Linux computer • up to 5 MHz • The combination of small form Linux 2.6 ready, ARM9 embedded computer Radstone Embedded Computing factor, considerable channel density, onboard • One 10/100 Ethernet • Four serial ports with Website: www.radstone.com signal conditioning, and high sample rate RS-232/422/485 • USB for wireless LAN and Model: ICS-1745 RSC No: 30257 makes the ICS-1745 suitable for demanding flash disk expansion • Internal SD socket for High-speed data acquisition system in a 3U applications where space, weight, and power flask disk • Internal 32 MB SDRAM • Built-in CompactPCI form factor • All channels feature are critical factors • Particularly well suited to unmanned and autonomous underwater vehicles performing littoral sonar and mine hunting applications

Data recorders Conduant Corporation Website: www.conduant.com Model: Amazon SATA Controller RSC No: 30355 Provides real-time recording and playback disk controller technology for the PCI bus • Sustained recording and playback at 400 MBps www.embedded-computing.com/rsc – guaranteed • Interface card to card via the PCI bus or direct to disk via Conduant’s modular mezzanine interface system • Power failure recovery support avoids unwanted data loss • Wrap mode for continuous recording with overwrite • Data forking for real-time playback while recording • Simultaneous streaming (read/write) for asynchronous playback while recording • Multichannel recording • Additional standard or custom mezzanine interface cards available upon request

Development platform MicroTech International Ltd. Website: www.microtech.com.pl/m/eng/index.asp Model: Spartan 3 board RSC No: 30341 Low-cost, small-dimension, high-performance board based on efficient Xilinx’s Spartan 3 -1000 chip • Features an open architecture – most pins available in the expansion connectors and can be used as general purpose pins for I/O device connection, A/D converters, and more • Can be used as a flexible ready part for application circuit development • Suitable for educational purposes, systems prototyping, and IP core development and testing • Measures 100 x 70 mm • Four layers PCB • MMC card slot onboard • RS-232 onboard for serial communications • JTAG onboard for FPGA configuration • 8x LED indicators for basic lab work • 3x 40-pin extensions connectors for 120 general purpose Spartan pins

Pinnacle Data Systems, Inc. Website: www.pinnacle.com Model: AdvancedTCA Products RSC No: 30254 Fully telco-grade PICMG 3.x compliant • Meets NEBS requirements including cooling, RSC# 44 @ www.embedded-computing.com/rsc noise, and shielding • A platform for building

44 / July 2006 Embedded Computing Design carrier-grade telecom applications supporting for prototypes, production, and custom orders CPU: VIA MARK 533 MHz/MARK 800 MHz • five-nines (99.999 percent) availability • High- • Design assistance, finishing, labeling, and System memory: 1x 144-pin SODIMM SDRAM density/high-performance computing environ- assembly • No minimum order • An unlimited socket support up to 512 MB • Ethernet: Dual ment • Entry level to advanced product choices variety of customized profiles • A choice of RTL8100C for 10/100 Mbps Ethernet • I/O inter- • Complete systems integration and support four standard profiles • Complies with PICMG face: 4x USB 1.1, 2x SATA, 1x LPT/FDD, 1x CFII, services • AC and DC power options 3.0 specifications • Unsurpassed EMI and RFI 1x RS-232, 1x RS-232/422/485, 1x PS/2 KB/MS shielding • Excellent heat dissipation • Any 1x IDE • Display: CRT integrated in VIA MARK size, any time, any quantity 533 • LVDS: 24-bit TTL/dual 18-bit LVDS inte- Development tools grated in VIA MARK • RoHS compliant Embedded Toolsmiths Industrial computers Website: www.etoolsmiths.com Software: Middleware Model: JTAGFLASHPro RSC No: 30356 IEI Technology USA Corp. www.embedded-computing.com/rsc JTAG flash programming hardware and soft- Website: www.usa.ieiworld.com Themis Computer ware for PowerPC, MIPS, XScale, and ARM Model: WAFER-MARK RSC No: 30369 Website: www.themis.com processors • Automatic flash device identi- An industrial computer that provides more Model: Quorum SLAs Software RSC No: 30345 fication • Fast programming speed • Easy to LAN, SATA, USB, and digital I/O functions Automated management software to monitor use • Supports all popular flash devices • with a low-power, fanless solution to expand and control distributed computing systems and Windows or Linux GUI or command line applications into more vertical markets • applications • Policy-based system performs

Enclosure Chomerics Website: www.chomerics.com Model: PREMIER RSC No: 30208 EMI shielding thermoplastic for volume production of electronics and related components • Greater than 85 dB shielding effectiveness • Low through resistance down to 30 mΩ highly conductive • High permeability (6.5) • High tensile strength and modulus • High flexural strength and modulus • Low density for weight reduction up to 75 percent • Recyclable, conforms to WEEE EoVL TCO • RoHS- and EPA-compliant and halogen free • Up to 105c RTI • Corrosion-free for long life • Lower total cost of ownership by eliminating secondary operations • Six-sigma processing • Waste elimination • Global supply • PREMIER shor- tens the supply chain, saving packaging and shipment costs to specialized coaters or suppliers and Work in Process (WIP)

Fabrics: PCI Express PLX Technology Website: www.plxtech.com Model: PCIe Bridges RSC No: 30210 Each device includes an integrated PCI Express PHY that provides 2.5 Gbps PCI Express lane speed • Each device’s PCI Express port is compliant with the PCI-SIG specification 1.0a with polarity reversal • PEX 8111 and 8114 support forward/reverse bridging while PEX 8114 also offers nontransparent bridging mode • PEX 8111’s 10 mm x 10 mm package is suitable for applications with limited board space and power budget • PEX 8311 offers direct connection to two industry-standard interconnect buses • PEX 8311 includes 2 DMA channels to offload CPU • PEX 8111 is available in a 144 PBGA and 161 Fine Pitch PBGA • PEX 8114 is available in a 256 PBGA • PEX 8311 is available in a 337 PBGA

Front-panel hardware XTech Website: www.xtech-outside.com Model: Face Plates-AMCs RSC No: 30216 Lighter than stainless steel • Fast turnaround RSC# 45 @ www.embedded-computing.com/rsc

Embedded Computing Design July 2006 / 45 automated closed loop management of Thermal management application QoS and allocation of resources for OpenSystems Publishing™ OpenSystems Publishing real-time systems • Automated monitoring and Degree Controls, Inc. management of SLAs and Application Quality Website: www.degreec.com of Service (AQoS) • Helps CIOs and real-time Model: ProntoFlow Control-ATCA RSC No: 30211 Advertising/Business Office resource and application provisioning • Re- Real-time, IPMI-based selection of several duces hardware requirements for high- fan control modes for synchronized, closed 30233 Jefferson Avenue availability systems and reduces the need for loop RPM control of fans • IPMI-based recon- St. Clair Shores, MI 48082 operator intervention • Supports a Manager- figuration of alarm thresholds and fan control Tel: 586-415-6500 n Fax: 586-415-4882 of-Managers architecture that allows a distri- curve set points, and access to fan speeds • Vice President Marketing & Sales buted system to be hierarchically connected Private, I2C-based communication bus between Patrick Hopper to a higher-level manager • Dynamic system fan trays for those chassis containing multiple [email protected] www.embedded-computing.com/rscconfigurations that can be changed on-the-fly fan trays • Fire response functionality to either • Device discovery for new hardware • Integra- reduce or increase fan speed, depending on Business Manager tion with third-party management systems chassis architecture needs • Multiple filter Karen Layman through CIM • Customizable Visualization Layer clog and maintenance schemes • Fan fail pre- with a Java API for OEM-specific look and feel diction and fan fail monitoring to ensure fan Communications Group tray replacement occurs as fan performance degrades Patrick Hopper [email protected] For more information enter the product’s RSC# at Christine Long www.embedded-computing.com/rsc Online Marketing Specialist [email protected]

Embedded and Test & Analysis Group Dennis Doyle Senior Account Manager [email protected] Page/RSC# Advertiser Product description Doug Cordier Account Manager 27 Acces I/O Products PC/104 Embedded Systems [email protected] 35 Adlink ETX and ETXexpress Modules Barbara Quinlan 45 Advantech SOM Modules Account Manager 20 Annapolis Micro FPGA Systems [email protected] 10 Arcom Intel XScale Based SBCs 2 Diamond Poseidon Epic SBC Military & Aerospace Group 44 EDT PCI Boards Tom Varcie 17 Embedded Planet Embedded Planet Professional Services Account Manager 48 IBASE OEM/ODM Services [email protected] 41 ICP GoPC-Mobile Andrea Stabile Advertising/Marketing Coordinator 9 ICS 600 MBytes/second Record and Playback [email protected] 47 Intel Embedded Designs 25 Jacyl Digital FPGA and Analog FPAA Regional Sales 6 Kontron ETXexpress-CD Jane Hayward 7 Lattice Semiconductor Soft Core 8-Bit Microcontroller Regional Manager – California 5 Microsys SBC1586, SBC1495, SBC2590 [email protected] 28 One Stop Systems CompactPCI Express Development Systems Phil Arndt 29 Performance Tech CPC5564 64-Bit AMD Opteron Single Board Computer Regional Manager – East Coast 13 Radian Thermal Analysis [email protected] 16 Technologic TS-7300 Richard Ayer Regional Manager – West Coast 12 Tews Embedded I/O Solutions [email protected] 38 Themis SPARC SBCs 36 Toronto Micro ECM401 International Sales 23 Toronto Micro Embedded Computer Solutions Stefan Baginski 34 Toronto Micro Micro-P3 European Bureau Chief 24 Tri-M 2300mA-hr NiMh Battery Module [email protected] 32 Tri-M EES-3611 Dan Aronovic 43 Versalogic Extended Life-Cycle Policy Account Manager – Israel [email protected] 31 VMETRO Vanguard Express 3 WinSystems PC/104, EBX, EPIC Reprints and PDFs Call the sales office: 586-415-6500

46 / July 2006 Embedded Computing Design OpenSystems Publishing™ OpenSystems Publishing

Advertising/Business Office 30233 Jefferson Avenue St. Clair Shores, MI 48082 Tel: 586-415-6500 n Fax: 586-415-4882 Vice President Marketing & Sales Patrick Hopper [email protected] Business Manager Karen Layman

Communications Group Patrick Hopper [email protected] Christine Long Online Marketing Specialist [email protected]

Embedded and Test & Analysis Group Dennis Doyle Senior Account Manager [email protected] Doug Cordier Account Manager [email protected] Barbara Quinlan Account Manager [email protected]

Military & Aerospace Group Tom Varcie Account Manager [email protected] Andrea Stabile Advertising/Marketing Coordinator [email protected]

Regional Sales Jane Hayward Regional Manager – California [email protected] Phil Arndt Regional Manager – East Coast [email protected] Richard Ayer Regional Manager – West Coast [email protected]

International Sales Stefan Baginski European Bureau Chief [email protected] Dan Aronovic Account Manager – Israel [email protected]

Reprints and PDFs RSC# 47 @ www.embedded-computing.com/rsc Call the sales office: 586-415-6500 RSC# 48 @ www.embedded-computing.com/rsc ATCA

Sign up to receive the next issue at www.radisys.com/go/radisys. ATCA’s BLAZING 10-GIGABIT ARCHITECTURE SPEED AVAILABLE TODAY

AMC O 1.800.950.0044 RadiSys is a Premier member of the Intel Communications Alliance, a CONVERGING ON V IP? 503.615.1100 global community of communications and embedded developers and COM EXPRESS www.radisys.com solutions providers committed to the development of modular, INTEL CORE DUO standards-based solutions based on Intel technologies. With well over a AFFORDABLE PERFORMANCE, IMPRESSIVE EFFICIENCY hundred members worldwide, the alliance is delivering economies of scale to the communications industry, accelerating the development of MICROWARE OS-9 REVIEW optimized, multi-vendor solutions based on industry standard LESS POWER, MORE PERFORMANCE technologies and Intel communications building blocks. RACKMOUNT SERVERS MEDICAL OEMs SEEK LOW RISK, LONG LIFE MOTHERBOARDS RadiSys Corporation 5445 NE Dawson Creek Drive THE SECRET ONLY EMBEDDED Hillsboro, OR 97124 PRODUCT DESIGNERS KNOW RADISYSRADISYS LETTER FROM FROM THE THE CTO CTO

INNOVATION DOESN’T HAPPEN in Just ONE PLACE

Welcome to the first edition of the RadiSys Product and Application Guide. We hope you enjoy the enclosed articles. Throughout this guide you will see evidence of RadiSys’ innovation. We believe all innovation happens because someone thinks differently. My team and I, the CTO Office and the Engineering Organization, spend time thinking like our customers, what they are building and what they need. We talk with them constantly. We even talk to their customers so we can understand how they want to use the products our customers will build. That puts us in a position to be aware—to know—what our customers will want.

Thinking ahead embraces innovation. If you start thinking of a product you will need three years from today and start implementing it a year from now, it will likely reach the market in the right time. Thinking in this Geor ge Shenoda manner puts us in the forefront of the industry providing products and CTO & VP of Engineering technologies well ahead of the competition such as the 10-Gigabit ATCA RadiSys Corporation Platform, which is the first product of its kind on the market providing our customers with a state-of-the-art high performance platform that can serve as a base for multiple network elements.

At RadiSys, innovation is finding how to make our products in more effective ways that make it easier for our customers to achieve time to market with high value. We strive to create products with higher performance, lower cost, smaller size—even better looking. The look of a product can make the difference in how effective it is by creating the proper perception of the product.

How we do it is simple. By creating requirements that are beyond the ordinary, beyond what is on the market, pushes engineers to think creatively.

We simultaneously take both a bottom-up approach and a top-down approach to defining, architecting and implementing our products. We see the product’s functions, features, architecture and performance from our customers perspective of networking and simultaneously look at the latest technology, the most advanced protocols, and the best silicon available to implement the products. We push the envelope by putting products in development today using silicon parts and software that are not on the market yet and we use common building blocks to maximize the utility of our products.

Innovation doesn’t happen in one place; it happens across the industry— and only with continual interaction. One good idea spawns another. It’s a regenerative process which creates a perpetual innovation. TABLETABLE OF OF CONTENTS CONTENTS RADISYS

ATCA 2 ATCA’S BLAZING 10-GIGABIT ARCHITECTURE SPEED AVAILABLE TODAY Utilizing a 10-Gigabit switching architecture lays the framework for a speedy platform.

AMC 6 CONVERGING ON VOIP? PLANNING & SELECTING A MEDIA GATEWAY Evaluate the risk and benefit of VoIP by balancing the right solution at the right value with the right connections to wireless and IP networks.

COM EXPRESS 8 INTEL CORE DUO AFFORDABLE PERFORMANCE AND IMPRESSIVE EFFICIENCY Announced this year, Intel® Core™ Duo is now appearing in embedded computers like the RadiSys Procelerant™ CE945GM.

MICROWARE OS-9 RTOS 11 HIGH-RELIABILITY, HIGH-PERFORMANCE EMBEDDED SYSTEM DESIGN Maximize performance and minimize power consumption with Microware OS-9.

RACKMOUNT SERVERS 13 MEDICAL OEMS SEEK LONG LIFE, LOW RISK High performance and product stability are critical attributes for embedded services used in medical imaging.

MOTHERBOARDS 15 THE SECRET ONLY EMBEDDED PRODUCT DESIGNERS KNOW Embedded single-board computers touch our daily lives as they automate, separate, sort, pack, track, stack and inspect our goods and our lives.

LOOK FOR THESE ICONS

APPLICATION NOTES WHITE PAPERS TIME SENSITIVE Quickly learn about Technology focused Get the latest information important trends, and information to provide you on new products and quickly market applications in next with access to our leading evolving issues that will help technology sectors. edge knowledge base for you make next step decisions. topics ranging from design to application considerations.

©2006 RadiSys Corporation. RadiSys is a registered trademark of RadiSys Corporation. Promentum and Procelerant are trademarks of RadiSys Corporation. *All other trademarks are the properties of their respective owners. 07-1322 0506 www.radisys.com | TC RADISYS ATCA

Once just an ancillary network control application for session setup, monitoring and management, ATCA (Advanced Telecom Computing Architecture) is moving into diverse applications. Combining ATCA with 10-Gigabit Ethernet brings into play new high-bandwidth, compute intensive solutions, including IMS (IP Multimedia Subsystem), radio network controllers, media gateways, and call servers.

With new managed platform and modular building blocks, TEMs (Telecommunication Equipment Manufacturers) can move ATCA into core traffic-bearing applications that demand high-bandwidth and traffic processing power. These new modular computing architectures deliver the highest speed I/O, switching and packet processing—up to 10-Gigabits— configured in a managed ATCA platform and for system applications that include CSCF (Call Session Control Function), application servers, edge routers, voice over IP, wireless network elements, security, and IP television.

| www.radisys.com ATCA RADISYS

MAKING THE BEST CHOICE While 10-Gigabit Ethernet switching platforms Staying with 1-gigabit switching architectures for are the future in terms of enabling multimedia, the next 12 to 14 months is a short-term choice broadband services, few vendors are marketing but all is not lost. If it is early in the development them. For TEMs, service providers and network cycle then there is still time to make the change. operators’ wanting to move up to a bandwidth- Looking at the market in which the application enhanced platform, the question is not if they will be deployed will help drive the correct should migrate but when. decision. The market conditions to take into consideration are: TIME TO MARKET Will the system be deployed into an Pressured to deliver products to market faster, environment with growing bandwidth TEMs and service providers want to turn on requirements (Wireless networks, as an services quickly and move high call volumes as example, have seen significant growth in soon as possible. Their solution is in a common the throughput delivered to subscribers.) platform that not only meets today’s needs, Will the system be required to interwork a but “future-proofs” their solution. 10-Gigabit number of different I/O flavours (i.e. ATM, switching fabrics provide the higher bandwidth OC-3/STM-1, Ethernet?) that platforms need to enable a wide variety Will the system be deployed in a test of configurations. market with expected rapid growth in the For example, the RadiSys® Promentum™ number of subscribers? SYS-6010 offers several options to meet the divergent needs of this next generation of APPLICATION NOTE applications. The SYS-6010 provides a 10-Gigabit Ethernet switch and control module, the highest bandwidth node AND connectivity, up-link and cross-link capability CSCF APPLICATION SERVERS and the highest processing density in each ENABLE MULTIMEDIA SERVICES FOR slot. AMC-based I/O and packet processing NETWORK OPERATORS provide modularity that allows for flexible The combination of the CSCF (Call Session Control expansion of both features and system density. Function) and application servers within the IMS Using modules for common platform design (IP Multimedia Subsystem) architecture provides provides the tools to address a diversity of network operators a way to create new multimedia requirements from each of the elements in a services on a low-cost, easy to manage, flexible, and common network platform. For example, a CSCF highly-integrated platform. Building applications on a solution needs a highly scalable distributed high-availability CSCF/application server platform lowers computing function, while a media server the economic risk, the unpredictability of consumer requires an abundance of storage. Burdening adoption and go-to-market risks... a CSCF solution with storage costs makes that solution more costly and uncompetitive. To download the complete application note visit: www.radisys.com/go/atca

www.radisys.com | RADISYS ATCA

Reduce Your Risk 10-Gigabit architectures are the best long- When there is a need to make a switching term solution because it offers the capacity to choice, and, for that matter, a platform choice, it aggressively market multimedia and broadband is imperative to involve vendors that can identify applications that can increase revenue and the strengths and weaknesses of a solution as market share. It also reduces the risk of the it applies to a specific application in a specific unpredictable consumer response by providing market. This enables the correct decision when the flexibility to test new services in smaller deploying next generation application and markets, learn what customers want and then platforms, and is the first step towards success. 2G/3G WIRELESS expand the capacity of the service. Aggregates wireless voice and data. Eric Gregory, Sr. Product Line Manager email: [email protected] E A C T I A O D ACCESS EDGE V ACCESS EDGE PRODUCT SPOTLIGHT: MGW IP MGW PROMENTUM™ ATCA SYS-6010 VOICE Converts voice to CALL VoIP and provides SERVER services in VoIP (call waiting). Radisys Promentum SYS-6010—The and reliable architecture complete with industry’s first 10-Gigabit managed comprehensive system management. The common platform and modular building Promentum building blocks are implemented GGSN blocks for next-generation wireless and in a modular fashion with seamless WIRELESS DATA wireline applications. interoperability. Modularity is made possible PDSN with Promentum building blocks and a suite of Converts from PROMENTUM Until now, ATCA has been targeted for use in wireless protocols to AMCs (Advanced Mezzanine Cards), which can IP and provides ATCA SYS-6010 ancillary network control applications such as be used to configure functions such as general connectivity to the session setup, monitoring and management. internet. purpose and network processing, input/output, INTERNET With the introduction of the RadiSys® and storage. In order to further increase their Promentum™ SYS-6010 managed platform velocity to market, RadiSys includes data path and new modular building blocks, TEMs APP and platform management software integrated (Telecommunications Equipment Manufacturers) SERVER APPLICATION into the Promentum product line. This will can now take advantage of ATCA in core traffic- SIP enable TEMs to focus on developing the higher Provides services bearing applications that require significantly like Push-to-Talk, value application layer rather then spending Instant Messaging, greater bandwidth and traffic processing precious resource developing basic protocols and Internet Gaming. capabilities. The Promentum SYS-6010 delivers and platform management solutions. the highest speed I/O, switching and packet E N processing capabilities to be implemented I L E R in a managed ATCA platform and can be I W WIRELINE configured for use in applications such as IMS (IP Multimedia Subsystem), Radio Network Aggregates Controllers, Media Gateways and Call Servers. B-RAS broadband lines.

The SYS-6010 is a fully integrated and validated managed platform designed to address high speed I/O and bandwidth intensive traffic- bearing applications implemented in a flexible

| www.radisys.com ATCA RADISYS

2G/3G WIRELESS

Aggregates wireless voice and data. E A C T I A O D ACCESS EDGE V ACCESS EDGE MGW IP MGW VOICE

Converts voice to CALL VoIP and provides SERVER services in VoIP (call waiting).

GGSN WIRELESS DATA PDSN Converts from wireless protocols to IP and provides connectivity to the internet. INTERNET

APP SERVER APPLICATION SIP Provides services like Push-to-Talk, Instant Messaging, and Internet Gaming. E N I L E R I

W WIRELINE

Aggregates B-RAS broadband lines.

ATCA IN A CUSTOMER’S NETWORK

The above diagram demonstrates the ATCA candidates spread across the network that can be implemented in your network.

www.radisys.com | RADISYS AMC

Broadband access and Internet protocol Platforms to Meet Your Customer’s Needs (IP) networks have nearly delivered the When planning solutions for your customers, help promise of information at everyone’s fingertips. them consider the business value of moving to But in our information rich society, we each VoIP and the way they will deploy it. For example: choose our own “communicator” according to our work- and life-style. It might be a Blackberry, Work with them to select a media gateway integrated cell phone, handheld device, laptop that is comprehensive, cost effective and bears no penalties for a fully loaded box or a desktop PC. And wherever we are, we also in the future. want all forms of media—data files, e-mail, voice, music, video—wirelessly poured into our personal Know their existing voice and data traffic and the effects on their network today, communicator. Pushing these multiple media including their current voice and data types through a common Internet pipe boosts usage patterns. bandwidth needs higher every year. Two of the Define what audio quality, reliability and big issues for service provider and enterprise uptime is acceptable for each customer. customers is connecting all these communicators and how to preserve the quality of service when Right-sizing VoIP Media Gateways for the connections go wireless. This is especially Enterprise and Service Providers true when they want VoIP (Voice-over Internet To evaluate the risks and benefits of VoIP for Protocol) to be wirelessly packetized and yet enterprise or service providers, balance the right maintain wireline quality. solution with the right value and with the right connections to wireless and IP networks. This More and more these companies want their value chain takes detailed planning and a deep employees to have rich-content media and they understanding of what you want to achieve. are considering VoIP to cut telecom costs. But they need to understand that what seems cost- Offer flexible configurations for your effective today can cost them dearly in the future. customer’s application When TEMs (Telecommunications Equipment A carrier-grade shelf and server platform meeting Manufactuurers) sell VoIP solutions to service 5x9 performance offers capacity to add more providers or enterprise customers, there are processors and edge cards. Both ATCA and several questions to ask so that you can select MicroTCA architectures offer this. A combination the right-sized, modular media gateway that of double-wide AMCs (Advanced Mezzanine makes it possible for your customers to minimize Cards) in either an ATCA or MicroTCA chassis is their risk before they leap into a VoIP project. also an alternative.

| www.radisys.com AMC RADISYS

Know the customer’s need to scale up or down Application call volume determines a customer’s approach. Deciding the configuration that suits the application’s market size and future scaling drives the decision. The market size and call volume influence the customer’s architecture choice, as well as the ability to scale up or down by adding new combinations of cards and chassis. Finding the right balance helps keep the overall system cost lower.

Provide cost-effective solutions for customer applications To address customer scaling, telecommunications APPLICATION NOTE equipment providers need a modular, flexible system that can scale call volumes from a few hundred to thousands. ATCA handles call volumes over 10,000. The advantage of ATCA Modular Media Gateways is the high density it enables—8000 calls or more per slot. However, the same advantage Enable VoIP and Other could become a drawback if the requirement is to scale in smaller steps, not to mention the Media Services: Make the Right Choice and Get to Market Faster reliability of such high call volume on a single slot. To meet the smaller densities, leveraging Every day more companies are concluding that the modularity of AMC/MicroTCA is perhaps the enabling their IT infrastructure with VoIP is a worthy right choice. Once fully defined, MicroTCA will investment with both hard and soft returns on handle volumes under 1,000 calls. A mix of ATCA investment. They seek to lower their telephone and AMC or MicroTCA and AMC standards offers expenses, while improving workgroup and team cost-effective options. Customers using ATCA or communications by using media gateways to deploy MicroTCA chassis with AMC cards can handle call a VoIP (Voice-over Internet Protocol) solution while volumes of 500 and 10,000 callers. With lower setting the stage for video services. By acquiring media call volumes, one-to-one redundancy may not be gateways that connect VoIP to wireless and Internet necessary to deliver your customers the service networks from the PSTN (Public Switched Telephone quality they need. Network), companies can integrate VoIP into their IT infrastructure. This allows companies to get more out Built on open-system ATCA and AMC standards, of their IT investment by turning to an application- a media gateway, like the Promentum platform ready platform that meets their needs today and in and building blocks, help your customers stage the future. their multimedia services and solutions and increase their company revenues. Enterprise and service providers are in a competitive market To download the complete application note visit: making it important for you to provide media www.radisys.com/go/amc gateway architectures that are cost-effective, robust, flexible and scalable—for their VoIP needs today and their video needs tomorrow.

V. Prasannan, Sr. Director ATCA Product Line Manager email: [email protected]

www.radisys.com | RADISYS COM EXPRESS

INTEL CORE DUO Affordable Performance Impressive Efficiency

And new COM Express Duality Pumps up Embedded Performance

RadiSys, a Premier member of the Intel® Communications Alliance, is helping customers take the next leap in microprocessor architecture with new Intel® Core™ Duo processors. The Intel Core Duo processors include two cores in a single processor that enable lower power and higher performance and responsiveness than thought possible with past processor generations.

The RadiSys Procelerant CE945GM is the first generation COM Express product to fully utilize the Intel Core Duo processor. COM Express applications such as medical imaging, test and measurement, gaming, entertainment and industrial automation demand low power platforms with high performance features. Battery life, small spaces with limited air-flow, and cooling fan noise are a critical component of embedded COM Express applications. RadiSys provides thermal engineering expertise, while Intel incorporates thermal reduction features into its dual-core processors, making for a thermally optimized platform.

The Intel Core Duo processor can operate at

| www.radisys.com COM EXPRESS RADISYS

very low voltages and minimizes clock and signal is used as a macro component and enables switching, resulting in lower power dissipation developers to focus on designing application in the active state. The processor can enable specific I/O, software, and differentiating the the chipset to power down with the processor user interface. This modular concept is gaining in low-frequency modes to further reduce in popularity as keeping up with the processor dissipation. Additional Intel Core Duo processor generation design can demand resources that take features include Dynamic Power coordination, away from the greater value of an application. which enables individual cores to dynamically COM Express ensures future generation upgrades transition to Halt, Stop Clock and Deep Sleep with a performance path that contains new Intel power management states and Intel® Advanced Core Duo embedded processors as soon as they Thermal Manager, which includes a new digital are released. Improve time to market and include temperature sensor and thermal monitor on each performance improvements in any low power, individual core located close to the hot spots for high performance application with the RadiSys enhanced accuracy at higher temperatures to COM Express product line, and start with the enable precise fan control. The additional power Intel Core Duo processor based COM Express management features enabled in the Intel Core module, the CE945GM. Duo processor are brought to life by the RadiSys Jennifer Zickel, Product Line Manager COM Express CE945GM product. email: [email protected] The Intel Core Duo processor on a CE945GM COM Express module brings breakthrough WHITE PAPER performance to processor intensive data crunching applications such as imaging and COM EXPRESS gaming. The two mobile optimized cores THE NEXT enable parallel threads or applications to be GENERATION COMPUTER ON executed on separate cores with dedicated CPU resources to enable multiple demanding MODULE STANDARD applications simultaneously. The dual core processor features a shared 2MB level-2 cache This white paper outlines Computer on Module advantages that enables dynamic cache allocation across and discusses the ripple effect movements from legacy both cores to enhance performance and reduce and parallel bus interfaces to high-speed differential serial under utilization. Advanced Branch Prediction, interconnects have had on standard form factors. The the most optimal algorithm to use, significantly concept of a Computer on Module, or COM, is not new reduces the number of mi-predicted branches within the embedded computer industry. Various COM and in turn, increases performance. The RadiSys solutions and implementations have been around for years, CE945GM module increases the performance but none ever took hold as a dominant or de facto standard capacity of the Intel Core Duo processor in within the embedded computer industry... the smallest size (95 x 125mm) module by Download the complete white paper visit: adding the key feature of dual channel memory. www.radisys.com/go/com-express The dual channel memory prevents a bottle- neck from occurring as the dual cores access memory; using single channel memory can slow board level performance from 13%-67% depending on the application.

Using a dual core COM Express module can ease the design complexity and development time associated with a high speed, complex processor design by splitting the CPU design from I/O and Intel, the Intel logo, Pentium, Celeron, Core Duo, and XScale are trademarks or registered form factor design. The COM Express module trademarks of Intel Corporation or its subsidiaries in the United States and other countries.

www.radisys.com | RADISYS COM EXPRESS

Procelerant CE Procelerant CR100 COM Express COM Express

® ® 2GHz Mobile Intel Pentium 4 Processor - M PICMG COM Express Revision 1.0 compliant and 1.5GHz Celeron® M combined with Intel® 915GM Express Chipset on COM Express Supports Type 1 and Type 2 basic form module factor modules Intel® ICH6M I/O hub FlexATX form factor 7.5” x 9” PICMG COM Express compliant One COM Express module interface Basic form factor (95mm x 125mm) Two 32-bit 33MHz 3.3V PCI slots, with 33Hz operation capable Type 2 COM Express pin-out

Procelerant CE945GM COM Express Intel® Core™ Duo T2500 2.0GHz and L2400 1.66GHz processors Intel® 945GM Express Chipset Intel® ICH7M Digital Home PICMG COM Express compliant Basic form factor (95 x 125mm) Type 2 COM Express pin-out )

e Broadcom BCM5789 1000BaseTX iz S

Ethernet controller tual

c One SODIMM Socket for up to 2GB A

( memory mm Flexible PCI Express Options 125 x 1*PCI Express x16 mm 5*PCI Express x1 (or) 1

- 95 PCI Express x4 and 1* PCI Express x1 ard o COM Express standard features B ss e 8 USB ports xpr 2 SATA ports 1 ATA100 port COM E Phoenix BIOS with ACPI 3.0 Power Management Win XP/ Win XP Embedded / Red Hat Desktop Linux, Win CE Optimized passive and active heatsinks available

10 | www.radisys.com MICROWARE OS-9 RADISYS

MICROWARE OS-9 REVIEW …

HIGHPERFORMANCE HIGHRELIABILITY If you’re developing high-performance, high-reliability embedded applications, EMBEDDED choosing the correct real time operating system (RTOS) is critical to your application’s success. More and more designers are realizing that SYSTEM DESIGN solving performance and reliability requirements with in-house solutions requires far more time and effort than they expected. Supporting and maintaining this in-house development becomes an unexpected burden on your engineering staff. Windows and Linux-based solutions are one possibility—if your application doesn’t require deterministic, sub-millisecond response times. Successful projects that A better approach is to consider using a commercial get to market first RTOS that provides the speed, reliability and Simple Component- Advanced Complex support your application requires. RadiSys Real-Time based Flexible Networking, Applications, Control Dynamic Multimedia, and Protocols, Microware OS-9™ provides this support with the Updates Management and Networks drivers, board-level solutions, development tools, middleware and support you need. You’ll have Java Language a complete solution, with OS kernel, software OS-9 Memory Protection Process-based System Security components, networking, graphics, power RTOS Power Management Plug-in Support management and development tools. Microware Threaded Kernel OS-9’s performance and reliability have been proven in thousands of products world-wide, to Developer Value including automotive, aerospace, communications, Application Complexity Missed deadlines and industrial automation and medical applications. cancelled projects

www.radisys.com | 1 1 RADISYS MICROWARE OS-9

Bullet-proof Your Applications with Modular Code You’ve probably heard the importance of TOP REASONS writing modular code—but does your RTOS 7 TO USE actually make that process easier? By focusing OS-9 on modules as the building blocks for your application, OS-9 increases the reliability and 1 2 Speed:3 OS-9 is one of the fastest RTOSes 4 5 available.6 maintainability of your system. Unlike other 7 8 9 RTOSes that have a monolithic approach, 1 2 3 Reliability and security: Modular architecture OS-9’s module-based environment ensures 4 5 6 and use of processor MMUs protect against 7 8 9 memory faults, viruses and application errors that a crash in one process or thread doesn’t that crash other RTOSes. bring down the rest of the system as well. OS-9 also adds an extra layer of safety by 1 2 3 Development environment: Debug processes 4 5 6 and dynamically load system components embedding CRC codes within the module, 7 8 9 (even device drivers) without rebooting. giving additional protection against viruses 1 2 3 and accidental code modification. This elegant 4 5 Optimize6 performance with built-in system 7 8 and9 application profiling. Generate exceptionally architecture also makes debug and maintenance efficient code with a compiler tuned easier—you can add, remove and replace individual specifically for OS-9. components (even device drivers!) while the 1 2 3 4 5 6 Popular 32-bit processor support: OS-9 system is on-line and in-use. And it allows for 7 8 9 supports PowerPC, 68K, x86, ARMv3/v4, automated configuration management for each Intel XScale® technology, MIPS3000/32/64, product build. SH-3, SH-4 and SH-4A. 1 2 3 Disks or No Disks 4 5 6 Standards compliance: Enhanced portability 7 8 9 with support for BSD-Sockets, IPv6 and POSIX Depending on your needs, you can specify OS-9 Threads. Port applications from Linux to OS-9 for a diskless environment without I/O, or a 1 2 in3 days not weeks. 4 5 6 full-system supporting serial, disk, Internet and 7 8 Fully9 Scalable: OS-9 supports the full range more. OS-9 is multi-user, supports module and of applications, from small and deeply file permissions and prevents user applications embedded to complete solutions with from corrupting system resources including the graphics and networking. kernel or device drivers. And the kernel is easily extendible at run-time for customization.

Mike Lottridge, Product Line Manager email: [email protected]

white paper by Canadian Space Agency A Selection Methodology for the RTOS Market “In past years, the market of Operating Systems (OS) has been quite active. One of those key markets is to support embedded real-time applications in which the OS must guarantee the timeliness as well as the correctness of the processing. Many OS claim to be Real-Time Operating Systems (RTOS), but often, it is only by reviewing the OS specifications or detailed information that one can truly identify the OS that enables real-time applications.” To download the complete white paper visit: www.radisys.com/go/os9

12 | www.radisys.com RACKMOUNT SERVERS RADISYS MEDICAL OEMS SEEK LONG LIFE, LOW RISK

For doctors to see inside our bodies, 10 years or more after they are put into service. medical imaging equipment needs vast Clearly, this is an application that demands more computational power to present detailed 2-D than just a typical “white-box” PC or server to (Two-Dimensional) and 3-D (Three-Dimensional) provide computing power. images. In the past, this might have required Product stability is critical for medical imaging a supercomputer. Thanks to Moore’s Law, that and also for other long-life embedded same processing power is now available in applications such as industrial automation and compact, rackmounted servers, such as the test and measurement systems. These systems RadiSys Procelerant™ Server family. These are complex and often require integration servers are used by medical manufacturers in a of components and subsystems from many number of imaging “modalities,” such as X-Ray, vendors. Typical PC lifetimes of 9-15 months MRI (Magnetic Resonance Imaging) and CT and frequent, often undocumented, component (Computed Tomography). changes can result in unexpected incompatibility Processing power requirements are continuing problems that will bring the manufacturing line to increase as new imaging instruments offer to a grinding halt. Costly re-engineering and increased resolution, four-dimensional viewing re-qualification is often required to resolve these (3-D images in motion) and combine images types of problems. from multiple modalities. The RadiSys Procelerant Rackmount Server family Five to 10 year lifespan uses carefully selected long-life parts, including Medical imaging and test instruments typically processors and chipsets on the Intel long-life have product lifetimes that stretch over several embedded road map, and well-documented years from initial design to production of the engineering change-control procedures to ensure last unit. These products often stay in use for consistent product performance.

RMS420-0945RB Long-life embedded server

RACKMOUNT SERVER — MichaelPCI and Reunert, PCI Express Product-line I/O Manager email: [email protected] Gigabit ethernet

Intel® Pentium® D or Celeron® Processor

Intel® 945G Express Chipset

www.radisys.com | 1 3 RADISYS RACKMOUNT SERVERS

Scalable Performance manufacturers want the flexibility to distinguish Medical imaging covers a wide range of their products. System designers can often performance and price points ranging from reduce cost and speed time to market by starting portable systems used in a doctor’s office to very with a standard RadiSys server and adding high performance systems filling entire rooms application-specific cards, selecting CPU speed in hospitals or imaging centers. As a result, and adding or deleting memory or drives to tailor there is no “one-size-fits-all” processing solution performance for a specific application. Custom for medical imaging. The RadiSys Procelerant faceplates for most Procelerant servers allow server family ranges from value products with a customers to add their own “look and feel” to a single Celeron® processor to high performance product. Extended operating temperature range products with multiple Xeon® processors. and low noise at normal operating temperatures Multiple Procelerant servers are used for the are added features that allow customers to highest performance imaging systems. differentiate their products.

Whether buried deep inside a large system or Michael Reunert, Sr. Marketing Manager externally mounted and visible to users, RadiSys email: [email protected] servers are an integral part of an OEM customer’s products. Whatever the approach, system

RadiSys offers regular Webinar: Role of Signaling Gateways seminars on the On-line June 1, 2006 latest technologies Global Comm influencing the Chicago, IL June 6- Jun 8 2006 embedded systems Webinar: Role of ATCA in IMS Development On-line July 11, 2006 design industry. Webinar: Building ATCA Systems & Software On-Line October 10, 2006

Check on-line for the most current information regarding our webinars, seminars and speaking engagements. Visit: www.radisys.com/go/events

14 | www.radisys.com MOTHERBOARDS RADISYS THE

SECRET ONLY EMBEDDED PRODUCT DESIGNERS [IT’S EVERYWHERE IN OUR LIVES] KNOW

Every day people step up to an ATM machine Top test and measurement and medical companies for fast cash. Parents get ultrasounds as depend on RadiSys to provide motherboards with part of pregnancy care. Surveillance cameras rigorous requirements. Their product engineers tape us when we buy gas. While different, each turn to RadiSys to provide long product life, robust of these machines has one thing in common. designs, customized features and to reduce their Pop their tops and you’ll find an embedded, field support costs. single-board computer. There’s a good chance it’s Long life and ruggedness make manufacturing, test one from RadiSys, like the new Intel® Pentium® and measurement and medical applications the processor-based OP945G microBTX in the RadiSys toughest of designs. They need both a stream of Procelerant™ Endura motherboard family. complex computational features, vibration analysis, RadiSys motherboards touch our lives daily. Only humidity testing and sometimes extensive thermal engineers developing embedded products know assessment for long-term and hard use. RadiSys exactly where. Manufacturing plants use RadiSys also gives product engineers a worldwide support motherboards powered by leading-edge Intel® organization to solve any problems their new processors to automate, separate, sort, pack, product designs may encounter—from thermal track, stack and inspect consumer goods on the analysis to selecting the right parts to assure years way to our homes. of use.

Product designers won’t lose sleep about product failures or upgrades years down the road. They can rest easy knowing RadiSys motherboards ease long- term field support and provide a smooth product upgrade path, because RadiSys provides a time-to- market product roadmap based on Intel’s long life, embedded chipsets and processors.

Peter Mitchell, Product Line Manager [email protected]

OP945G microBTX with an Intel® Pentium® 4 processor See chart on page 17 for specifications. www.radisys.com | 1 5 RADISYS MOTHERBOARDS

TP945GM mini-ITX with an Intel® Core™ Duo processor

The Endura TP945GM mini-ITX is a high performance, low power, small form factor motherboard for use with the new Intel® Core™ Duo processor. The board is highly integrated with multi-media features and a comprehensive set of I/O interfaces.

EM945G microATX with an Intel® Pentium® 4 processor

The Endura EM945G microATX is a high performance motherboard for use with Intel®’s Pentium® 4 and Celeron® processors. The integrated GMA950 graphics controller and a PCI Express graphics slot provide a flexible choice of video solutions.

16 | www.radisys.com ENDURA PRODUCT CHART MOTHERBOARDS RADISYS

BG845G SH845GV LS855 AB915GM BY915GV KP915GV EM945G RB945G OP945G TP945GM

Form Factor microATX ATX microATX FlexATX microATX ATX microATX ATX microBTX Mini-ITX

Size 9.6” x 9.6” 12” x 9.6” 9.6” x 9.6” 9” x 7.5” 9.6” x 9.6” 12” x 9.6” 9.6” x 9.6” 12” x 9.6” 10.5” x 10.4” 6.7” x 6.7” Intel Chipset 845GV 845GV 855GME 915GM 915GV 915GV 945G 945G 945G 945GM Intel Processor Pentium 4 Pentium 4 Pentium M Pentium M Pentium 4 Pentium 4 Pentium 4 Pentium 4 Pentium 4 Core Duo Celeron Celeron Celeron M Celeron M Celeron D Celeron D Celeron D Celeron D Celeron D Celeron M Socket 478-pin PGA 478-pin PGA 479-pin PGA 479-pin PGA LGA775 LGA775 LGA775 LGA775 LGA775 478-pin PGA Max CPU Speed 2.8GHz 2.8GHz 1.8GHz 2GHz 3.4GHz 3.4GHz 3.4GHz+ 3.4GHz+ 3.4GHz+ 2.0GHz+ Max Memory 2GB 2GB 2GB 4GB 4GB 4GB 4GB 4GB 4GB 4GB FSB 533MHz 533MHz 400MHz 533MHz 800MHz 800MHz 800MHz 800MHz 800MHz 667MHz PCI Slots 3 6 3 2 2 3 2 4 2 Mini-PCI x1 PCI Express Slots None None None 2 1 2 1 2 1 1 Graphics Slot AGP 4X ADD AGP 4X None ADD2 ADD2 x16 PCIe x16 PCIe x16 PCIe x16 PCIe VGA Extreme 1 Extreme 1 Extreme 2 GMA900 GMA900 GMA900 GMA950 GMA950 GMA950 GMA950 Ethernet 10/100, GbE 10/100 10/100 10/100, GbE 10/100, GbE GbE 10/100, GbE 10/100, GbE 10/100, GbE GbE #LAN Ports 1 or 2 1 1 1 1 or 2 1 1 or 2 1 1 or 2 1 or 2 Audio AC97 AC97 AC97 HDA HDA HDA HDA HDA HDA HDA 7.1 DVI Use ADD Use ADD Use ADD Dual DVI Use ADD2 Use ADD2 Use MEC Use MEC Use MEC Use MEC LVDS Use ADD Use ADD 24-bit LVDS 18-bit LVDS Use ADD2 Use ADD2 Use MEC Use MEC Use MEC 18-bit LVDS GPIO 13-bit 13-bit 13-bit 13-bit 13-bit 13-bit 13-bit 13-bit 13-bit 13-bit Flash Memory None CompactFlash None MM/SD None None None None None CompactFlash Serial Ports 2 2 2 None 2 2 2 2 2 4 Parallel Port 1 1 1 None 1 1 1 1 1 0 IEEE 1394b None None None Build Option None 1394b None Build Option Build Option None Watchdog(s) 1 1 1 1 1 1 1 1 1 1 System Management Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Hard Disk ATA/100 ATA/100 ATA/100 SATA SATA SATA SATA 300 SATA 300 SATA 300 SATA CD-ROM ATA/100 ATA/100 ATA/100 ATA/100 ATA/100 ATA/100 ATA/100 ATA/100 ATA/100 ATA/100 Floppy Disk Yes Yes Yes No Yes Yes Yes Yes Yes No Rear USB 2.0 4 4 4 4 4 4 4 4 4 4 Internal USB 2.0 2 2 2 3 or 4 4 4 4 4 4 4 PCI Riser Card Ext. Yes No Yes No Yes No Yes No No No

BG845G SH845GV LS855 AB915GM BY915GV KP915GV EM945G RB945G OP945G TP945GM

Last Updated: 2006-05-24 ATCA

Sign up to receive the next issue at www.radisys.com/go/radisys. ATCA’s BLAZING 10-GIGABIT ARCHITECTURE SPEED AVAILABLE TODAY