Intel® Cloud Builders Guide Intel® Xeon® Processor-based Servers Cloud Design and Deployment on Intel® Platforms: Oracle

Intel® Cloud Builders Guide: Cloud Design and Deployment on Intel® Platforms–Oracle

Cloud Design and Deployment on Intel® Platforms: Oracle

AUDIENCE AND PURPOSE This reference architecture outlines an enterprise cloud infrastructure deployment using Oracle’s complete stack with servers, storage, network, and software components. The Oracle* Optimized Solution for Enterprise Cloud Infrastructure makes use of pre- configured Oracle VM virtual machines, that contain optimized configurations for running Intel® Xeon® Processor 5500 Series applications on Oracle’s 6000 Modular Systems with Intel® Xeon® processor based blade servers and Oracle’s Sun ZFS Storage Appliances. Using the contents of this Intel® Xeon® Processor 5600 Series paper, which includes detailed scenarios and screen shots, should significantly reduce the learning curve for building and operating your first cloud computing infrastructure. IT organizations typically spend many weeks to plan, architect, and deploy a multi- vendor solution: the process is not only time consuming, but also error-prone, making it hard to get the best value from the investment. This document offers a much simpler approach that speeds deployment and reduces risk. It is a single-vendor solution for the entire hardware and software stack and can be deployed in hours or days rather than weeks. Because the creation and operation of a cloud can require integration and customization to existing IT infrastructure and business requirements, it is not expected that this paper can be used “as-is.” For example, adaptation to an existing network and identification of management requirements are out of scope for this paper. Therefore, it is expected that the user of this paper will make adjustments to the design to meet specific customer requirements. This paper is assumed to be a starting point for that journey. www.oracle.com/goto/cloud

June 2011 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Table of Contents

Executive Summary...... 3 Introduction...... 3 Oracle* VM...... 4 Oracle VM Server for x86...... 4 Oracle VM Manager...... 4 Oracle VM Templates ...... 4 Sun Blade 6000 Modular Systems...... 5 Sun Blade X6270 M2 and X6275 M2 10GbE Server Modules...... 5 Oracle Sun Blade 6000 Ethernet Switched NEM 24p 10GbE...... 5 Sun ZFS Storage Appliances...... 5 Fast I/0 Throughput for Virtualized Environments...... 5 Flexibility Through Configuration Choices...... 6 Intel® Cloud Builders Usage Scenarios...... 6 Deployment Key Concepts...... 6 Test Bed Blueprint Overview...... 11 Test Bed Design Considerations ...... 11 Hardware Description...... 11 Virtual Machine Configurations...... 11 Software Description...... 12 Technical Review...... 13 Intel Cloud Builders Use Case Details...... 13 Assumptions...... 14 Tests and Results...... 14 Conclusion...... 46 Additional Considerations...... 47 Storage ...... 47 Scalability...... 47 Networking...... 48 Processor Description...... 49 Parallelism...... 49 Intel® Turbo Boost Technology...... 49 Intel® Hyper-Threading Technology...... 49 Intel® QuickPath Technology...... 49 Intel® Intelligent Power Technology...... 49 Intel® Virtualization Technology...... 49 Intel® AES New Instructions (Intel® AES-NI) Technology...... 50 Intel Web Resources...... 50 Oracle Web Resources...... 51 Glossary...... 52

2 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Executive Summary Oracle Enterprise Cloud Infrastructure Builders scenarios have been implemented This document describes how the investments. Through a lifecycle services by Oracle EMEA (Europe, Middle East, Oracle* Optimized Solution for Enterprise approach, Oracle experts can help Africa) and Intel. architect customized solutions that reduce Cloud Infrastructure can be utilized to Oracle VM Templates are used to simplify deployment risk and ensure operational provide an enterprise cloud computing and accelerate deployment of the Oracle readiness. deployment. It covers deployment of Optimized Solution software stack. software, hardware, storage, and network For the Intel Cloud Builders test cases, Introduction components and is intended to serve as a pre-installed VM template versions of practical guide to help organizations get The Oracle Optimized Solution for * and Oracle VM Manager up and running quickly while maximizing Enterprise Cloud Infrastructure is used were used. These were deployed on the potential benefits of Oracle VM for to build the Intel® Cloud Builders test Oracle VM Server, which runs directly quickly building a cloud deployment using configuration with Oracle’s complete on Oracle’s Sun Blade servers. Best Oracle’s Sun Blade servers based on Intel® infrastructure stack with hardware and practices for optimizing the environment Xeon® processors and Sun ZFS Storage software components (Figure 1). The are built into the templates, which define Appliances. optimized solution makes use of pre- configurations for all components of the configured virtual machines that contain Oracle Optimized Solution for Enterprise software stack. Oracle Secure Global optimized configurations for running Cloud Infrastructure provides a complete, Desktop is utilized for remote desktop applications on Oracle’s Sun Blade Modular integrated solution to dramatically access. Systems with Intel Xeon processors, reduce the time required to deploy Oracle’s Sun ZFS Storage Appliances, and Infrastructure-as-a-Service clouds, and Intel® architecture based Storage Servers. deliver lower total cost of ownership and The hardware components have been improved productivity for enterprises and tested with Oracle VM by the respective cloud service providers. Oracle Advanced product teams, and the Intel Cloud Customer Services offers a suite of services that maximizes the value of

Figure 1: Oracle* Optimized Solution for Enterprise Cloud Infrastructure provides a complete hardware and software stack that can accelerate deployment and reduce the risk of errors

3 Oracle Optimized Solution for Enterprise Cloud Infrastructure

The Oracle Optimized Solution for (XE), Standard Edition (SE), Enterprise Oracle VM Templates Enterprise Cloud Infrastructure is Edition (EE), and Real Application Clusters Oracle VM Server provides the ability designed, deployed, and operated as (RAC) are supported as the management to rapidly and easily deploy a pre-built, a single system allowing customers to repository. Oracle VM Manager manages pre-configured, pre-patched guest virtual deploy applications in hours instead of the virtual machines running on Oracle machine (or multiple machines depending days. Tests show drastic reductions in VM Server through the Oracle VM Agent. on the application). The guest VM can virtual machine deployment times, with For Intel Cloud Builders tests we have contain a complete software solution significantly lower cost of acquisition, installed it as a VM, using the Oracle VM along with the operating system and support and operating costs. Manager Template. related software infrastructure. These Complementary to the Oracle Although not discussed in the Oracle guest VMs, called Oracle VM Templates, Optimized Solution for Enterprise Optimized Solution for Enterprise Cloud are available from Oracle’s E-Delivery Cloud Infrastructure, users can then Infrastructure, Oracle VM Servers can Website and are ready to download and simplify and accelerate application or alternatively be managed using the run. Already configured for production database deployments using pre-defined, Oracle Enterprise Manager Grid Control use, Oracle VM Templates can save users customizable Oracle VM Templates Oracle VM Management Pack, which is days or weeks learning to install and which contain ready-to-run software licensed separately from Oracle VM. The configure a sophisticated product such as configurations that can be up and running Oracle VM Management Pack provides Siebel CRM or Oracle Enterprise Manager in hours as opposed to weeks. Oracle VM management functionality in Grid Control. Instead, users can simply a similar manner to Oracle VM Manager, download and start the VM to begin using Oracle VM but is integrated into Oracle Enterprise it right away. To fully understand the rationale behind Manager Grid Control. This enables Within the downloadable Oracle VM the described best practices and their management and monitoring of Oracle Templates, Oracle software is configured implications, it is important to review VM environments alongside all other in the same manner as if the software some of the key concepts and components Oracle and third-party products managed had been installed and patched manually. of the Oracle VM solution. For x86 by Enterprise Manager. This integrated The package and patch inventories are servers, Oracle VM consists of Oracle VM management approach offers deeper completely standard and up-to-date so Server for x86 and Oracle VM Manager. insights into the health, performance, that no changes to operations procedures and configuration of the hardware and are required to maintain the instances Oracle VM Server for x86 software stack. over time. Accordingly, Oracle VM Oracle VM Server for x86 installs directly Oracle VM software stack can be deployed Templates can also be fully customized on server hardware including Intel Xeon in private and public clouds. This paper post-install and then re-saved as “golden processors that support PAE (Physical discusses how Oracle VM software stack image” Templates in Oracle VM. Address Extension) and does not require is used to create a private cloud. Oracle Oracle VM Templates can serve as a a host operating system. It requires HVM customers can also run their production user’s enterprise deployment standard (Hardware Virtual Machine) support (e.g. workloads in public clouds using Oracle to minimize risks and variation across Intel® Virtualization Technology) on the VM. The public cloud support allows multiple instance deployments. An up-to- underlying hardware platform in order customers to add or shed resources date list of available Oracle VM Templates to run HVM guests. The Oracle VM Agent as needed, paying only for resources can be found on the Oracle VM Templates for server management is installed with used and freeing up scarce engineering Website. Oracle VM Server. resources from running technology Oracle VM Manager infrastructure. Amazon EC2 now uses Oracle VM Server to run Oracle software Oracle VM Manager is a -technology- in the public cloud. In September 2010 based management server. It uses an Amazon Web Services* and Oracle Oracle database as its management announced Certification and Support repository, which can be installed either of Oracle Applications, Middleware, and on the management server or a separate Databases on Amazon Elastic Compute server. Oracle Database Express Edition Cloud using Oracle VM.

4 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Sun Blade 6000 Modular Systems infrastructures that require the high the virtualized architecture of the Oracle Sun Blade 6000 Modular Systems density of a dual-node blade and support Optimized Solution for Enterprise Cloud from Oracle provide an open modular for 10 Gigabit Ethernet to reduce I/O Infrastructure. architecture that is optimized for bottlenecks. The Sun Blade X6275 Key benefits include: performance, efficiency, and density. With M2 10GbE server module is Oracle’s a choice of server modules including the highest compute density blade server • Simplified network topology with latest Intel Xeon processors, organizations with support for a total of 24 CPU cores embedded 10GbE switching, eliminating can select the platforms that best and up to 192GB of main memory in the need for external switching match their applications or existing a single dual-node blade. Combining • Reduced costs and simplified infrastructure, without worrying about the Sun Blade X6275 M2 10GbE with management with 4:1 cable reduction vendor lock-in. The blade servers also the high performance, low latency Sun and simple scalability to four chassis support a choice of operating systems as Blade 6000 Ethernet Switched Network • Seamless integration into existing data well as industry-standard PCI Express* Express Module (NEM) 24p 10GbE delivers centers with the industry's highest (PCIe) I/O modules, expandable storage, the required bandwidth to eliminate density uplinks transparent networking, and consolidated potential network bottlenecks that can be management. experienced in virtualized environments. Sun ZFS Storage Appliances Sun ZFS Storage Appliances provide all Oracle blades are optimized to run Oracle Oracle Sun Blade 6000 Ethernet the benefits of unified storage in an Switched NEM 24p 10GbE VM, Oracle Linux, and *. easy-to-use appliance package to help The result is a modular IT infrastructure The Oracle Sun Blade 6000 Ethernet organizations dramatically simplify their architecture that serves the needs of the Switched NEM 24p 10GbE simplifies data storage deployment and management data center and the goals of the business center deployments while simultaneously while also reducing costs. These while protecting existing investments into delivering the highest attainable appliances change the economics of the future. throughput with 10GbE switching storage by using a high-performance technology. With its combination of Hybrid Storage Pool (HSP) architecture Sun Blade X6270 M2 and X6275 M2 advanced cut-through switching, novel use 10GbE Server Modules that combines the performance of QSFP for uplinks, and line rate quality characteristics of SSDs with high-capacity Within the Sun Blade 6000 Modular of service, the switched NEM 24p provides hard disk drives (HDDs). The SSDs provide System Chassis, the recommended the ideal solution for virtualization and read and write cache, enabling higher blade server modules for the Oracle network storage. The solution scales to performance than traditional storage Optimized Solution for Enterprise Cloud four chassis, offering 75 percent fewer architectures at up to 35 percent less Infrastructure are the Sun Blade X6270 cables and eliminating the need for cost. M2 and the dual-node X6275 M2 10 expensive external switching. Gigabit Ethernet (10GbE) server modules. Oracle’s Unified Storage systems also In the Oracle Optimized Solution for provide unmatched simplicity and Oracle's Sun Blade X6270 M2 server Enterprise Cloud Infrastructure, the Sun ease-of-use through an intuitive and module is the best fit for mid-range Blade 6000 Modular System is equipped powerful browser user interface (BUI) for virtualization workloads that require a with a redundant pair of switched NEM simplified management. Revolutionary balanced blade. It supports up to 144GB 24p interconnects. Each is dedicated business analytics functionality allows of main memory and up to 12 CPU to one head of the Sun ZFS Storage administrators to quickly diagnose and cores with the high-speed six-core Intel Appliance cluster. This solution leverages resolve performance issues in production Xeon processor 5600 series. The Sun the bandwidth of 10GbE while ensuring systems. Blade X6270 M2 server module offers no single points of failure. It also offers balanced performance and superior numerous unused 10GbE uplinks per NEM Fast I/O Throughput for Virtualized energy efficiency while reducing the for seamless integration into the data Environments total cost and complexity of a virtualized center. Sun ZFS Storage Appliances support infrastructure. many concurrent I/O threads due to an The high performance and reduced cabling innovative architecture that leverages Sun Blade X6275 M2 10GbE server offered by the Sun Blade 6000 Ethernet SSDs, multiple CPU cores, and 10 Gigabit modules are ideal for large virtualization Switched NEM 24p 10GbE make it ideal for

5 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Ethernet connectivity. This high level of physical servers and virtual machines. may also be securely Live Migrated or I/O throughput enables Sun ZFS Storage The Intel Oracle Bid Support Centre uses automatically started or restarted across Appliances to support more virtual Oracle VM Server and Oracle Secure Global any servers in the pool. In the Intel Cloud machines without sacrificing service Desktop to deliver a cloud computing Builders test lab, Oracle’s Sun ZFS Storage levels. environment to partners and customers Appliances offer managed storage for across EMEA. Oracle Secure Global Oracle VM Servers to share storage via While the appliances support many Desktop remote access to the Intel Cloud iSCSI and OCFS2. different storage I/O protocols including Builders test environment is illustrated as Common Internet File System (CIFS), VMs are associated with a given server an example – this enables Oracle VMs to Internet Small Computer System Interface in the pool. This association is made be accessed from anywhere. (iSCSI), InfiniBand* (IB) Technology, and dynamically at power-on and is based Fibre Channel (FC), NFS over Ethernet is Please refer to the Technical Review on load balancing algorithms or on a recommended for attaching the storage section for details of how the use cases user-defined list of named servers called in the Oracle Optimized Solution for have been implemented using Oracle VM. the Preferred Server list. The Preferred Enterprise Cloud Infrastructure. This is Server list identifies which servers are Key Deployment Concepts due the extreme simplicity of NFS and the to be used as a host for specific VMs. fact that it scales well on Sun ZFS Storage From a deployment perspective, multiple When VMs are powered-off and not Appliances. Oracle VM Servers are grouped into running, they are not associated with server pools as shown in Figure 2. Every any particular physical server. Powered- Flexibility Through Configuration server in a given pool has access to shared off VMs simply reside in the shared pool Choices storage, which can be NFS, SAN (Fibre storage. To meet a variety of needs for capacity, Channel), or iSCSI storage. This allows VMs price, and performance, Sun ZFS Storage associated with the pool to start and run As a result of this architecture, VMs can Appliances are available in different on any physical server within the pool. easily start-up, power-off, migrate, and/ configurations that include high Typically the server chosen is the server or restart without being blocked by the availability (HA) cluster configurations that has the most resources available, failure of any individual server or by the for the Sun ZFS Storage Appliance 7320, or a server that closely matches the failure of multiple servers, as long as 7420, and 7720 system models. In a resource requirements of the VM. Given there are adequate resources in the pool clustered configuration, the head node the uniform access to shared storage to support the requirements for all VMs to resources failover to the surviving node mounted under the /OVS directory in run concurrently. in the event of node crash, thus helping which all resources will reside, VMs to ensure near continuous availability. All systems come bundled with the same software including data protocols, replication, compression, and Oracle Solaris Dynamic Tracing (DTrace) analytics software for system troubleshooting and performance optimization.

Intel® Cloud Builders Usage Scenarios Intel Cloud Builders test cases explore key aspects in the management and usage of a cloud computing environment. Oracle has implemented the Intel Cloud Builders test cases using Oracle VM Server to illustrate how Oracle VM can support these requirements. Oracle VM Manager was used throughout to manage the Figure 2: Oracle* VM Server Architecture

6 Oracle Optimized Solution for Enterprise Cloud Infrastructure

As shown in Figure 2, there are different 2. Utility Server support normal and peak workloads as roles of Oracle VM servers in the pool. In each server pool, there can be one well as what types of workloads are These roles, which are defined below, are or multiple servers designated with the appropriate to share to the pool or server. implemented by the agent running on the Utility Server role. Utility Servers are Workload profiles should be considered Oracle VM server. Multiple roles can be responsible for performing resource- in addition to how predictable or combined in a single server, but there is intensive copy or transfer activities in unpredictable the workloads may be. only one Server Pool Master in the server the pool. For example, cloning activities, There are also some significant similarities pool at any given time. creating VMs from Oracle VM Templates, between server pool planning and physical 1. Server Pool Master or saving existing VMs as templates all server planning with regards to node involve either copying an existing VM In each server pool, there is exactly one (physical server) size versus overall pool image or moving an image from one Server Pool Master at any given time. It size. For example, in some cases it is directory to another. Depending on the coordinates a number of the activities of better to have relatively fewer but larger size of the files involved, this activity can the pool, particularly when that action servers in a pool. In other cases, a greater be resource-intensive. It may be desirable requires coordination across multiple number of relatively small servers or to off-load this activity from servers servers. blades are a better fit. Both deployments that are hosting production VMs so as to may provide the same aggregate CPU, This includes such key activities as minimize or eliminate any service level memory, storage, and bandwidth but the coordinating Secure Live Migration actions impact. Multiple servers in a pool can be implications of the deployment in a pool as well as Guest VM HA (high availability) designated as Utility Servers to provide can be different. auto-restart, and power-on actions better load balancing and availability as amongst others. described later in this paper. For Oracle VM 2.2 and later, each server pool must have its own shared storage All of the Oracle VM agents in the pool 3. Virtual Machine Server resources that can be accessed by Oracle communicate directly with the Server Pool Virtual Machine Servers are simply VM servers within the same pool. A Master, which, in turn, communicates with servers that host VMs. At least one separate server pool must have its own the Oracle VM Manager. This architecture server in a pool must be a Virtual Machine separate shared storage. provides a number of benefits including Server. Depending on total server capacity 5. High Availability Planning high management scalability in large requirements, it may or may not be environments as well as higher availability separate from the Server Pool Master Oracle VM provides the following features at the Oracle VM Manager instance level and/or the Utility Server(s) in the pool as to provide maximum up time for VMs because functionality is distributed and described later in this document. running in server pools: isolated to minimize the impact of any • Guest VM HA — Auto-restart on server single failure. For example, a management 4. Server Pool Planning or VM failure. server outage does not prevent the ability There are a large number of • Secure Live Migration — Move VMs off of the pool master(s) to complete Secure considerations when planning the virtual of servers that are undergoing planned Live Migration tasks or to automatically infrastructure and one size does not fit all. maintenance. fail over/restart failed VMs. This section provides some considerations • Automatic pool load balancing on and guidelines to help develop a plan that Similarly, a pool master outage on one VM start-up — At VM power-on, an is well suited to an organization’s unique pool does not affect the operation of algorithm dynamically assigns the requirements. another pool. By default, the first server host server for the VM in order to load added to a pool is assigned the role of It may be helpful to think of the server balance, but also to avoid a down server the Server Pool Master. With Oracle VM pool as if it were one big server with blocking VM start. 2.2, the Server Pool Master can perform an aggregate amount of CPU, memory, Figure 3 highlights how these availability auto-failover when a server pool is HA storage, and network bandwidth. As features are typically deployed. The enabled and a virtual IP address is set for such, planning for deploying VMs into a features should be used to collectively the server pool. pool is much like planning for a server consolidation. It involves deciding how much aggregate capacity is needed to

7 Oracle Optimized Solution for Enterprise Cloud Infrastructure

maximize up time of guest VMs running • Plan to use the Secure Live Migration Utility Servers are automatically load in a pool. The details of these features feature to migrate VMs in support of balancing. If one Utility Server fails, the and capabilities are the subject of planned events like server maintenance remaining Utility Server(s) can continue another white paper titled, “Oracle VM – to prevent any service outage. without any disruption of utility Creating & Maintaining a Highly Available • Plan for enough excess capacity in services (VM cloning, VM import, Create Environment for Guest VMs”. aggregate across the pool to support VM from Templates, etc.). The following section summarizes best running all VMs to an appropriate 7. Pool Capacity, Performance, and practices and considerations for planning service level even when one or more Scalability server pools. servers in the pool are out of service. Up to 32 nodes are supported within a Capacity planning for a server pool is 6. Recommendations for Deploying server pool. similar to capacity planning for a physical server. However, the following additional Highly Available Server Pools • Plan for multiple server pools when considerations are also important when The following steps are recommended to there is a need for more than 32 planning capacity for a server pool: achieve high availability: physical servers. • Enable HA and set up virtual IP (VIP) Plan extra capacity to support Guest HA/ • Enable the “High Availability (HA)” for the server pool to take advantage auto-restart. There should be sufficient option at both the server pool level of the Server Pool Master auto-failover capacity to support hosting additional and individual VM level to ensure VMs to minimize risk and thus maximize VMs on relatively fewer machines in the are automatically restarted after an availability of the management services. event that one or more of the servers unplanned failure. If the Server Pool fails and its VMs end up being restarted Master fails in a High Availability set up • Configure two or more dedicated and on the remaining, healthy servers, if only of Oracle VM, another Oracle VM Server load-balancing Utility Servers to provide temporarily. is automatically selected from the continuity of service in the event of server pool to act as the Server Pool a utility server outage. There can be Plan for extra capacity to support Live VM Master. multiple Utility Servers in a pool and Migration during planned events. When performing maintenance on a server (or servers) in the pool, Live Migration allows administrators to migrate VM(s) to another server in the pool without interrupting service. To take advantage of this capability, there should be enough excess capacity in aggregate across the pool so that a server can be taken offline (after migrating its VMs) without inappropriately impacting service levels. 8. Determining How Many Servers or VMs a Pool Should Contain Before creating a server pool, you need to consider how many physical Oracle VM Servers are to be included in the server pool, and what functions(s) each physical Oracle VM Server is to perform. The more guest virtual machines you run in the server pool, the more resources these guest virtual machines will consume, and so the more physical Oracle VM servers may be required to provide sufficient Figure 3: High availability helps minimize downtime due to planned or resources for the server pool. unplanned events

8 Oracle Optimized Solution for Enterprise Cloud Infrastructure

A server pool is scalable. If you find a Of course, the answers to these storage 11. Service Level Support Strategy server pool does not have sufficient questions depend on the I/O environment. Sometimes service level objectives resources, such as CPU or memory, Is the application I/O intensive? What is the may dictate that there is enough to run the virtual machines, you can average size of an I/O request? planned excess capacity to support expand the server pool by adding more Determining a realistic number of storage normal service levels even if all the Virtual Machine Servers. The number of nodes in a storage cluster requires VMs peak at the same time. This is the servers or VMs that are ideal for a pool considering similar questions: most conservative option, but also the depends on a number of factors that can most expensive since it requires extra vary greatly between data centers and • How much I/O will each storage cluster hardware that may not be utilized deployments. There is no single correct node support? much of the time. answer to this question, but there are • How much I/O will each server generate to/ several factors that should influence from the clustered file system? Another alternative is to plan for the average load and accept any such decisions. Some considerations are • How many servers will be accessing the performance hit based on resource described in the remaining points below. clustered file system? contention if there is too much 9. Storage Topologies, Performance, 10. Workload Profile of a peak. This certainly reduces and Implementation Is the workload flat and stable or variable hardware expense, but may not Oracle VM pools require that all servers and peaky? The ideal VM has relatively low provide acceptable service levels if the in a pool have shared access to the same utilization and is very flat and stable with workloads are too unpredictable or are storage so that VMs can be moved around minimal peaks. These types of VMs can often mission critical. As a result, many data easily. This means that server pools be very tightly consolidated, with a large centers plan their capacity to support must use shared storage such as NFS, number of VMs per server. Since they are very some percentage of the aggregate OCFS2 (Oracle’s Cluster File System), or predictable they require little excess capacity peak load. For example, they may SAN (iSCSI or FC) storage. In the Oracle or headroom to be available to accommodate plan for 40-60 percent of the peak Optimized Solution for Enterprise Cloud unexpected peaks. above average. This is often a good Infrastructure, Oracle’s Sun ZFS Storage compromise between meeting service Appliances are used to provide NFS or The next best scenario is when consolidating levels and having reasonable hardware OCFS2 shared storage. Both the physical multiple VMs that may have peaks, but the utilization. However, it clearly depends make-up of the storage devices and the peaks are very predictable in both timing on how critical service levels are for scalability of the file systems used will and magnitude. For instance, some VMs may the given workloads. dictate how many servers are practical contain applications that always peak at 12. Dynamic Server Pools for a given shared-storage pool without the end of the week or end of the month. If adversely affecting I/O performance. those VMs could be consolidated with other Understanding that the server pool VMs that peak at exactly the opposite time is dynamic, keep in mind that VMs are When evaluating how many servers – workloads that peak at the beginning of not associated with any one physical can share a given storage topology, the the week or month– they can potentially be server until they are powered-on and following questions should be considered: packed fairly tightly to maximize the number placed on a server. Typically they are • How much I/O will each server generate of VMs per server and per pool. placed on the server that currently has the most memory free. Based on and can the throughput and latency The worst case scenario is when the VMs are HA events or Live Migrations, VMs needs be accommodated through the highly variable in load and in timing. In this can be moved around within the pool. designated NIC or HBA ports? situation, it is likely that a comparatively large Thus a given VM can end-up sharing a • How much I/O can the storage device or amount of extra headroom will be needed on server with any other VM in the pool devices support? the servers. Thus fewer VMs will be able to be unless policies are implemented to • What are the HA needs (bonding or accommodated per server and per pool. restrict which servers can host which multipath)? VMs. This means that the capacity • Are there any application requirements plan really needs to be at the pool for directly accessed storage? level, not the individual server level. So it may be best to consider keeping

9 Oracle Optimized Solution for Enterprise Cloud Infrastructure

highly volatile VMs in their own pool or components of the same application important to make sure there is enough restricting them to a subset of the pool stack are not hosted on the same server capacity to fail-over and to generally where a relatively large amount of excess in order to maximize availability. These support peak loads. It is also important capacity can be maintained for handling policies are respected for all actions to have the option to Live Migrate either unpredictable peaks. within the pool, including Live Migrations, component (VM) to another server to Power-On, and HA/Auto-Restart. One support server maintenance without any Conversely, highly predictable VMs should consideration in using Preferred Server application down time. In this simplistic be restricted to a separate pool where policies is that it can have the effect of case, there would need to be a minimum the resources can be very tightly planned increasing the number of servers required of four nodes since each of the two VMs for high utilization without the need for in the server pool to support HA and Live must have two servers available. The first much excess capacity. Migration. server would act as the primary server 13. Server Affinity Policies for running the first VM and another Additionally, for any server in the pool, the server would be required for migrating Consider the impact of VM placement. Virtualization Server role can be removed or re-starting the first VM in the event of Oracle VM allows administrators to to prevent that server from hosting any a failure. The second VM would similarly implement policies on an individual VM VMs, although this is typically only done require its own two servers for primary basis to restrict which servers are allowed to create dedicated Utility Servers. To and migration/re-start. If the restriction to host that particular VM. Preferred give a simplistic example, when deploying that both VMs can never operate on the Server lists are used to implement these a VM application with two clustered same server at the same time is lifted, policies. They contain a list of explicitly components, it is a good idea to create a the two VMs could be hosted on just two named servers that are allowed to host Preferred Server policy for each VM. Such servers and still have HA and migration specific guest VMs within a server pool. a policy assures that both of the nodes support. Preferred Server policies are typically would never be on the same physical implemented to assure that two server at the same time. Yet it is still

Figure 4: Oracle* Optimized Solution for Enterprise Cloud Infrastructure

10 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Test Bed Blueprint Overview to be configured in a similar way so that with innovations that can help boost the service level agreements can be met performance, increase consolidation Test Bed Design Considerations during failovers or migration. ratios, and enable servers of different This section shows an Oracle VM generations to be combined in the same Each Sun Blade server module must be deployment scenario using Sun Blade virtualized server pool, improving virtual fitted with a Sun Intel 32599 Dual 10GbE 6000 and Oracle’s Sun ZFS Storage machine failover, load balancing, and PCIe 2.0 Fabric Expansion Module (FEM). Appliances. It’s a typical enterprise disaster recovery capabilities, making it an The FEM supplies each server with two deployment model using high-end Intel excellent choice for virtualization based shared 10GbE ports. The two 10GbE Xeon processor based servers and cloud deployments. storage. interfaces are used by VMs to access NFS services on the Oracle’s Sun ZFS For the main Intel Cloud Builders tests we The Oracle Optimized Solution for Storage Appliances. Two 10GbE interfaces used the Sun ZFS Storage 7420 Appliance Enterprise Cloud Infrastructure illustrated supply plenty of bandwidth and cable to provide shared storage for Oracle in Figure 4 offers high performance, aggregation for virtual machine data. For VM, enabling High Availability and Live availability, reliability, and manageability NFS and iSCSI access, 10GbE ports are also Migration functionality for the VMs. and is specifically intended for used. database virtual machines. While these We implemented equivalent infrastructure requirements can be met in numerous Intel® Cloud Builders Lab Hardware at both the Intel Oracle Bid Support Centre (Intel, Ireland) and the Oracle ways, this reference architecture For the Intel Cloud Builders lab we utilized Solution Centre (Oracle, Scotland). At implements an optimal approach. The five Sun Blade X6270 M2 server modules, Oracle we used the OSC’s existing shared Intel Cloud Builders test lab environment plus a single Sun ZFS Storage 7420 Oracle Secure Global Desktop deployment, closely follows these guidelines as Appliance for Oracle VM shared storage, whereas at the BSC we installed our explained in detail below. hosted at the Oracle Solution Centre at own. In a production environment this Oracle, Scotland. The following sections detail the would either be replicated or configured server components, virtual machine The five Sun Blade X6270 M2 server as high availability VMs depending on allocation, storage configuration, and modules were each configured with two requirements. In both cases our software project allocation with Oracle’s Sun ZFS Intel® Xeon® processor X5680, 72GB installation and configuration work was Storage Appliance storage and network memory, 2 x 300GB 10K RPM internal done remotely, using the browser based configuration. SAS-2 disks. administration consoles of the blade server chassis, and SGD. Hardware Description The Oracle VM Pool Master and Utility Both blade-based and rack-mounted Server were assigned to the first Sun The Intel Cloud Builders lab environment Oracle Sun x86 servers using Intel Blade X6270 M2 server module in the Sun has 1Gb but not 10Gb networking, and Xeon processors can be used to run the Blade 6000 chassis, and all five Oracle Sun five blade servers. Oracle VM infrastructure in a large, high Blade X6270 M2 server modules were availability configuration. As an example, used as Oracle VM Server Virtual Machine Virtual Machine Configurations Oracle’s Sun Blade 6000 Modular system Servers, hosting VMs. For each virtual machine, the number of virtual CPUs and memory can be with 10 Sun Blade X6270 M2 or 5 Sun Oracle’s Sun Blade 6000 modular configured to address the specific Blade X6275 M2 server modules can system is ideal for virtualized application requirements of the applications that are be configured to run the Oracle VM environments and application deployed. In the lab setup, a minimum of software. Each server is attached to a infrastructure products. Powered by the 4GB memory and two virtual CPUs are clustered Intel® Storage Server that is six-core Intel Xeon processor 5600 series, configured for each of the middleware accessed using NFS, or via FC/iSCSI LUN. they leverage the flexible I/O architecture and application virtual machines. For the This approach enables additional features of the Sun Blade 6000 chassis, making it virtual machine that hosts the database, such as live migration, high availability, the easiest blade system to deploy. Built 16GB of memory and four virtual CPUs are distributed resources scheduling to be on Intel® microarchitecture, codenamed allocated. Note that the number of virtual used for the VMs in the server pool. Sets Nehalem the Intel® Xeon processor CPUs can be tuned in a live system. of servers that are identified as part of 5600 series expands the benefits of a single server pool are recommended virtualization beyond consolidation

11 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Choosing the number of virtual machines Table 1 shows the recommended reads) benefit from more disks during per VM server as well as the type configuration of four distinct storage scenarios when there is a read miss and number of network links are also pools and the purpose of each pool. from the cache. critical considerations for a successful Further details about the recommended • Pool-2 — One or more projects are deployment. For example, for certain pools are provided below. created for storing the various I/O loads over a single 1GbE link, the Each storage head is set to be active for middleware components. Projects network could be saturated first before two pools. This enables both load-sharing might be dedicated to items such as either the server or the storage. In such and high availability for all storage pools. Web server, SOA binaries, Tlogs, admin cases, adding more links or moving to In the event of a failure in HEAD-1, HEAD- binaries and so on. one or more 10GbE links may provide 2 will take over ownership of pool-0 and • Pool-3 — One or more projects are a better performance. If the CPUs are pool-1 as well as continuing to serve the created for storing the various pegged running multiple virtual machines clients. application binaries and configurations on the same host, then either add for the Oracle E-Business Suite, Oracle additional servers to the server pool • Pool-0 — One Oracle VM project is Siebel, Oracle Peoplesoft and so on. and live migrate the virtual machines to created in pool-0 to store Oracle VM The shares (file systems) created under balance the Pool, or move the Oracle VM storage repositories. This is accessed these projects are mounted from the infrastructure from the midrange server from the dom0 of the Oracle VM server. various virtual machines. For security to an enterprise class server. • Pool-1 — One project is created to share purposes, access to certain projects and the ORACLE_HOME across the various For high availability and optimal file systems can be restricted to specific database instances. Additionally, performance an Intel architecture based clients (virtual machines). one project per database is created. Sun ZFS Storage 7420 Appliance is used Compared to the rest of the pools, as the storage platform. The servers are a greater number of disks are to be multi-core and the number of CPU cores allocated for this pool. This is because dictates the number of VMs that can be the random reads (db sequential assigned to each server.

Storage Configuration for High Performance with Oracle VM Pool Name Raid SUN ZFS System / VM Purpose configuration Storage 7420 Cluster Head

pool-0 RAID-Z2 HEAD-1 Dom0 For storing the Oracle VM storage repository. This pool is accessed from domain 0 and contains the OS images from which the virtual machines are launched.

pool-1 MIRRORED HEAD-1 Virtual machines For storing database files and database binaries. that run the One or more projects are created to cater to each database database instance.

pool-2 MIRRORED or HEAD-2 Virtual machines For storing the middleware components, including RAID-Z2 that run the Oracle binaries and configuration files. Fusion Middleware

For storing the application layer components – pool-3 MIRRORED or HEAD-2 Virtual machines including the binaries, configuration files. RAID-Z2 that run the Oracle Applications

Table 1

12 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Software Description Server and Oracle Secure Global Desktop The following software was installed in the Intel Cloud Builders test lab environment: can meet the real world tasks required to deploy a scalable private cloud. The use Technical Review cases were completed by using standard Oracle products without modification or Intel® Cloud Builders Use Case Details extension: Oracle implemented the Intel Cloud Builders use cases to demonstrate how Oracle VM • Create users, user group for business unit, and business unit administrator Intel® Cloud Builders Test Lab Environment Installed Software accounts System Component Software Component • Register and configure business unit Virtualization Platform Oracle VM Server 2.2.1 x86-64 resources (inc. physical servers) • Download here • Create new users and view users • Oracle VM is server virtualization software • Configure and use virtual machine that fully supports both Oracle and non-Oracle templates applications and delivers more efficient • View and modify virtual machine performance network settings VM & Deployment Management Oracle VM Manager Template v2.2 • Create services by virtual machine • Download here creation • Installed as a VM for Intel Cloud Builders, • Scale out service by cloning virtual but can be installed natively (includes Oracle machines Express as configuration DB) • Add additional physical servers to the business unit cluster Remote Access Oracle Secure Global Desktop 4.6 • Permit user access to virtual machines • Download here (previously Sun Secure Global through portal Desktop) • Monitor the state of services and virtual • Oracle Secure Global Desktop is utilized machines to allow remote desktop access to the lab infrastructure, Oracle VM Manager and the • Remove service by terminating and virtual machines themselves optionally deleting virtual machines • Automated failover to deal with VM Operating System for Oracle Linux* 5u5 – using Oracle VM Template hardware failure Compute Nodes OEL5u5 x86-64 PVM • Handling requests to scale deployed • Download here services • Important note: Other Operating Systems can • Decommissioning services and virtual be installed machines • Distributed application (data analysis) Storage Node Sun ZFS Storage 7420 Appliance version test by using multiple VMs '2010.08-17.1.1,1-1,16'.

Table 2

13 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Assumptions 1. Oracle VM components (Server and Manager) have been installed correctly 2. Any required Oracle VM Templates have been downloaded and imported using Oracle VM Manager (Oracle Linux PVM template is used in the example) 3. An Oracle VM administrator is available to create the Business Unit Admin accounts as illustrated in the test cases (or the Oracle VM administrator credentials are available)

Tests and Results Intel® Cloud Builders Use Case Summary Create Business Unit Admin Accounts Login to the Oracle VM Manager Web interface using the URL set when Oracle VM Manager was installed. e.g. https://ovmm.#####.com:8888/OVS default user “admin”, password set during install process of OVM Manager. You can also access by the IP of your Oracle VM Manager instance, instead of by name. Click Administration tab, then User. Click Create to create a new user. Complete new user details for each required BU admin user.

Create User Group for Business Unit Users We can also create a user group, BU-1, for our business unit and put our new BUadmin users in it.

14 Oracle Optimized Solution for Enterprise Cloud Infrastructure

15 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Register/Configure BU Cluster Resources Now, logging in as one of the BUadmin accounts just created, BUadmin2, we can see that the BU admin can access the physical server resources of the server pool, P1, which has been assigned to the BU. From here servers can be added to, or removed from the pool. Servers can be shutdown, rebooted, or put into maintenance mode. In the screen below, you can see that bscir2x33 is running additional components, Server Pool Master and Utility Server, to support the OVM Server Pool. In our lab five blades have been added to pool P1. Servers can be added and removed as required. The BUadmins can create new VMs, add and remove CPU cores, and virtual local and shared disks to the VMs as required. VM settings including network config, scheduling priority can be defined, VMs may be reallocated, live migrated to other servers, set for HA, cloned, etc.

16 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Create New Users/View Users Both BUadmin1 and BUadmin2 can create new users as necessary. Here we create four new users to illustrate, one at a time, from Administration > Users. After creation the user list—filtered with “c%” in the search field— looks like this:

17 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Configure Virtual Machine Templates The BU admins can create new VMs, either by importing existing Virtual machines from elsewhere, or by creating new VMs from Oracle VM templates. VM templates for Oracle products may be downloaded directly from Oracle, or may be produced by developers to build a local library. They allow new VM instances to be created easily by admins from a standardised VM configuration (template). In this case we have produced a new VM BU1-VM1 from the Oracle Enterprise Linux VM template that has been imported into our Oracle VM installation. Note that we assigned the new VM instance to group BU-1 when we created it, making it available to users in the BU-1 group, such as csmith.

BU admins can import templates and make them available to users of a server pool.

18 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Our end users can login and see the VMs that we created for the BU, BU-1. Here user csmith sees the VM created by the BU admins.

Configure Network BU admins can see the status of current VMs, modify servers, and reconfigure network settings for the physical servers in their pool(s) and their VMs. Here we select the BU1-VM1 VM.

19 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Clicking on the network tab for the VM shows current network details for the VM. Network interfaces may be added, deleted, and modified for the VM:

Here we change the Xen* network bridge for example. Additional networks can be added as necessary.

We can change the number of virtual CPUs (cores) assigned to the VM, relative scheduling priority compared with other VMs , scheduling cap, and whether the VM is configured for HA or not, amongst other settings. Cost management facilities for VMs are planned for future release. IP settings (including static, DHCP, gateway, DNS, etc.) can be configured to be set at first startup of the VM by definition within VM template scripts, or may be modified within the VM itself.

20 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Create Service Instance (instantiate VMs) BU admins can examine available VM templates, import, edit, activate, and deactivate others.

Here we create a new VM from a template by clicking “Create Virtual Machine”, including specifying create from template, choose a server pool that the VM can run on, the template that we want to use, etc. We set a name for the new VM and a password for connecting by VNC to the VM console and we select a network bridge from the interfaces available in the VM.

21 Oracle Optimized Solution for Enterprise Cloud Infrastructure

22 Oracle Optimized Solution for Enterprise Cloud Infrastructure

After confirmation, creation of the VM image then starts.

Later we can see that the VM was created successfully by the status, or by examining its logs.

23 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Additional virtual disks can also be created for the VM, or shared virtual disks attached/detached as required.

Create Multiple Virtual Machines (scale out service) Logged in as user BUadmin1, we start by selecting the VM that was created in earlier, and selecting cloning, then a prefix name for the clones.

24 Oracle Optimized Solution for Enterprise Cloud Infrastructure

After hitting Confirm, we see the clone VMs being created.

25 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Add Bare Metal Capacity to Existing BU Cluster Here we show the addition of server bscir2x37 to the existing cluster. The server is first installed with a vanilla install of Oracle VM Server. Then we use Oracle VM Manager to add it into the cluster – it has actually already been added here, but this makes no difference in illustrating the process. Starting from the Servers tab, press Add Server. Complete the details of the server to be added, and test the connection to the OVM server agent.

26 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Clicking on the server link shows available RAM, CPUs, and other info relating to the new server, plus the VMs that are running on it.

User Access to Virtual Machines Using Web Portal Oracle VM Manager can be deployed with https enabled on a public server. We have used Oracle Secure Global Desktop to enable remote access to our hosted Oracle VM environment, including Oracle VM Manager and the VMs themselves. This was installed on a separate server, but we could alternatively have installed the Oracle Secure Global Desktop template for Oracle VM, giving us a pre-installed instance. The SGD login is accessed from the public IP or registered name of its host server.

27 Oracle Optimized Solution for Enterprise Cloud Infrastructure

28 Oracle Optimized Solution for Enterprise Cloud Infrastructure

We just select the default “My Desktop” to access an Oracle Enterprise Linux User Desktop on the Oracle Secure Global Desktop server.

From there we can login to the OVM Manager as normal, through a browser. An SGD administrator can create panel shortcuts for direct access to desktop or terminal sessions on other VMs or physical machines.

29 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Monitor State of a Service Oracle VM users can login to the portal and view the status of the VMs and servers to which they have been granted access. They can create new VMs and disk quota within the limits of what is permitted by the servers to which they have been granted access. Oracle VM monitors the status of all VMs for which High Availability is enabled and will automatically restart these on available servers within the pool, if they should fail. Also the logs of the individual VMs can be viewed through Oracle VM Manager, showing the status of VMs over time. Extended monitoring facilities are planned for future release.

30 Oracle Optimized Solution for Enterprise Cloud Infrastructure

31 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Terminate VMs (remove service) Services may be removed by shutting down and removing the appropriate VMs. Note that Oracle PaaS (Platform as a Service) offerings will offer service level administration on Oracle VM. Below we see the current status of the hosted VMs. We have clicked Power Off for one of the VMs of the BU1-VM2-SO service. Once powered off, we can select to delete the VM instance. Note the checkbox to delete the VMs files from the Oracle VM shared storage.

32 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Balance Compute Resource Utilization Oracle VM automatically allocates VMs to servers within a pool of servers based on their availability and compute usage at allocation time. Oracle VM will only allocate VM workloads to servers within a pool that have sufficient resources. If insufficient resources are available on one preferred VM then it will choose another, either by auto-allocation, or from within a list of preferred servers. It is possible to assign “preferred” server(s) for VM startup. Oracle VM will try to use these servers to start the VM if they are available and have sufficient resources. In the example below we try to start VM BU1-VM2-SO-0000. We can see that the VM has been configured with three preferred servers. The BU admin has effectively defined that one of these should be used to run the VM.

33 Oracle Optimized Solution for Enterprise Cloud Infrastructure

We click to Power On BU1-VM2-SO-0000, then we see that the VM has now started on physical server bscir2x36.

34 Oracle Optimized Solution for Enterprise Cloud Infrastructure

If at another time there is not enough free RAM on bscir2x36 to run our VM, Oracle VM chooses another server from the list of preferred servers for the VM, and starts it on bscir2x35 instead. Note that VMs can be live migrated from one physical server to another without loss of service, using Oracle VM Manager.

35 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Create Multiple Virtual Machines (scale out service) In this example we have created a VM from a template, and then we choose to clone it to produce a scaled out service.

36 Oracle Optimized Solution for Enterprise Cloud Infrastructure

37 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Remove Virtual Machines (scale down service) The BU admin is able to stop and then delete virtual machines.

38 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Failure Scenario (Oracle VMs can be configured for HA, also live migration) We can show that a physical server can fail and the BU cluster remains operational. Oracle VM will restart VMs on alternate physical servers. It can also live migrate running VMs to alternate physical servers without loss of service. Oracle VM HA relies on having shared storage configured, which was done as part of the Oracle VM installation for the lab environment. We then need to enable HA for the Server Pool. The Server Pools > Edit Server Pool > Check button confirms whether the installation has been configured to support HA. If so then we just need to check the box to enable HA. We then enable HA for each VM that requires it. In a server failure scenario Oracle VM detects that the server is no longer available and after a short while restarts the VMs that were on it, choosing alternative server(s) from the server pool.

Service Change Request/Scale Service We can login as the BU admin and add additional storage and compute resources. We can add new physical servers to existing pools or create new pools of new servers. We can add new virtual disks and networks. After creation of the virtual disk it can be formatted and made available by the guest OS.

39 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Decommission Services BU admins can decommission deployed services by removing VMs. Here we delete a VM after powering off, similarly for all VMs that we wish to decommission. We check the box to physically remove the VM files from Oracle VM.

40 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Distributed Application (Data Better performance would be expected /OVS/running_pool. We did not use Analysis) Test Case by the specification of SSD devices for shared storage for these tests, but Hadoop* is becoming increasingly storage, 10Gb Ethernet between nodes, instead allowed each OVM Server to use attractive to a broad spectrum of and direct utilization of storage devices storage on its own local physical host organizations as a framework for running rather than the default file-mapped (local disks). This option was chosen computational workloads to process big storage, plus other potential optimizations as being closest to the requirement of data across large clusters of servers. which could be applied. Hadoop, where we prefer local storage for each node. Cloud customers can use private and The main purpose of this test is to public clouds to process their big data with show that Hadoop can be successfully • The standard OEL5 firewall was minimal investment by using a pay as you implemented on Oracle VM, and additional disabled on all VMs for the purpose go approach. We tested Hadoop TeraSort tuning of its performance is outside the of the test. (Note that some tests and Word Count benchmarks in our test scope of the Intel Cloud Builders test. may require changes to the config bed to show the feasibility of using virtual of the firewall on the physical server environments to run Hadoop applications. Additional comments on the use of Solid hosting each VM; do not disable the State Drives with Hadoop can be found in firewall on any server that has Internet For more information on Hadoop, please the Storage section. connectivity.) refer to: • All VMs were given static IPs and • http://hadoop.apache.org Hadoop Platform Configuration: configured to utilise the local DNS • http://hadoop.apache.org/common/ We tested Hadoop by using a four node server and gateway. (Changes can be docs/r0.20.2/index.html cluster. Each node is implemented by using made by modifying the network config an Oracle VM instance. For the Hadoop directly, or through “setup” at the OEL5 Background & VM portability: tests we chose not to use the Oracle’s command line.) We initially ran these Hadoop tests Sun ZFS Storage Appliance, but instead • Each VM was given two additional using Intel processor based blade server allowed each blade server to use its own virtual disks by adding local disks to hardware at the Intel Oracle Bid Support local disks, as this was considered more each VM through Oracle VM Manager; Centre at Intel in Ireland. We then copied appropriate for the Hadoop use cases resulting devices are visible under / a single VM from this environment to the which require local storage. proc/partitions. These virtual disks are Intel Cloud Builders test lab environment The following shows the test platform actually represented as files on the at Oracle, Scotland, where Intel processor configuration steps that need to be done actual storage devices of the physical based Sun Blade servers were used as to prepare the cluster to run the Hadoop host system – this is the standard detailed in the hardware description. We test cases: approach in Oracle VM. then cloned this single VM to provide the • Each virtual disk was then formatted additional required instances and modified • Four blade servers each have Oracle using fdisk to create a new partition the hostnames, static IPs, gateway, and VM server 2.2.1 installed; Oracle VM and then formatted as ext3 using mkfs. DNS settings accordingly. As a result Manager 2.2 was installed as a VM on ext3. we did not have to re-install the extra the first node. software components on each VM, saving • Each blade server has a single • Each disk was then configured to time and effort during the test process. Oracle Enterprise Linux 5u5 x86-64 automount by editing /etc/fstab in each The VMs ran without modification on both Paravirtualized Virtual Machine running VM. environments, showing the migration and on it (PVM). • A test user, “nurcan” was created on compatibility of Oracle VM instances on • In OVM Manager pools P1 - P4 were each VM, for the test scripts. This different brands of Intel Xeon processor created to allow each Oracle VM user had no password. (This was done based hardware. Server to use its own local storage, but to simplify test execution in a lab managed from a single OVM Manager environment, but should not be done in Note that no attempt was made to instance. Each pool contained one any live deployment.) optimize the Oracle VM installation for physical server. • SSH config on each VM was modified to Hadoop performance. • Each VM uses standard Oracle VM file based storage for its system disk, and therefore has its own directory under

41 Oracle Optimized Solution for Enterprise Cloud Infrastructure

allow the test scripts to SSH to each VM System (HDFS) system by formatting captured during some of the test steps as user nurcan, by adding “AllowUsers namenode described above. nurcan” to /etc/ssh/sshd_config. 6. Started Hadoop servers by using “PermitEmpty Passwords yes” was Hadoop “start-all.sh” script also added to allow the test script to run. (This should not be set on any VM 7. Executed steps 8-16 to test Hadoop or server in a live deployment, or on TeraSort benchmark any server that has connection to the 8. Generated test data for the Internet.) benchmark The platform configuration steps above 9. Started sar and iostat agents on each should be executed in an automated node fashion on a cloud portal based on the user’s cluster specifications. In our 10. Ran the benchmark test cases, we executed this custom 11. Stopped sar and iostat agents on configuration manually. each node 12. Created performance plots that Hadoop Test Software Stack: shows CPU, memory, disk, and We installed the following software on network utilizations for each node every node in the test cluster (except when VMs were cloned): 13. Collected performance data from all nodes • Jdk-6u23-linux-x64 (open source) 14. Created cluster performance metrics • Hadoop-0.20.2 (open source) that shows average CPU and disk • Python-3.1.2 (open source) utilization for the cluster • GnuPlot-4.4.2 (open source) 15. Created cluster performance metrics • Intel® Benchmark Install and Test Tool that shows disk and network (Intel tool) throughput for the cluster

Hadoop Configuration and Test Case 16. Created cluster performance metrics Execution: that shows device I/O rate and total The following describes Hadoop software memory utilizations for the cluster configuration and Hadoop test case 17. Repeated steps 8-16 to test Hadoop execution steps: Word Count Benchmark 1. A four node Hadoop cluster is 18. Stopped Hadoop servers by using created. One of the nodes is Hadoop “stop-all.sh” script designated as master node. We used The following are the screenshots the other three nodes as datanodes 2. Configured Hadoop to use two virtual disks to store HDFS data and map- reduce files on each node 3. Configured Hadoop to use the installed java home 4. Configured Hadoop nodes by using ulimit command to increase the limits for the user resources 5. Initialized Hadoop Distributed File

42 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Figure 5: Hadoop* Installation Directory on the Oracle* Cluster Nodes

Figure 6: Format Hadoop* Namenode

43 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Figure 7: Start Hadoop* Servers

Figure 8: Generate Hadoop* Test Data

44 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Figure 9: Run TeraSort Benchmark

Figure 10: Run Hadoop* to Display DFS Directories

45 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Figure 11: Stop Hadoop* Servers

Conclusion The entire software stack is pre- computing environment that is accessible Most IT organizations don't want to spend configured, pre-optimized, and ready- to our partners, customers, and our own weeks building a multi-vendor virtualized to-run on the recommended hardware staff from anywhere. It is built on Oracle infrastructure. Yet until now, they've had platforms. Because all the components VM Server, Oracle VM VirtualBox, and little choice because running IT without have already been tested and integrated Oracle Secure Global Desktop. together and are supported together, virtualization is too costly. The Oracle Oracle Optimized Solution for Enterprise there is less risk of deployment delays Optimized Solution for Enterprise Cloud Cloud Infrastructure also simplifies the or production downtime. With the ability Infrastructure offers the means to greatly support structure. Oracle Advanced to add compute, storage and networking simplify deployment and management Customer Services offers a suite of resources independently, the system of a virtualized cloud infrastructure services that maximizes the value of your can be scaled in multiple dimensions to while also reducing risk. Organizations Oracle Enterprise Cloud infrastructure match evolving changes in demand for can take full advantage of the hardware investments. Through a lifecycle applications. and software components in the Oracle services approach, Oracle experts can Optimized Solution for Enterprise Cloud The Intel® Cloud Builders Usage Scenarios help architect customized solutions Infrastructure, thereby realizing greater section described in this paper shows how that reduce deployment risk and ensure business benefits from their virtualized Oracle Optimized Solution for Enterprise operational readiness. There is no need infrastructure. Cloud Infrastructure can be deployed to negotiate multiple support contracts Oracle's innovative technology and with configuration of Oracle VM and with different vendors, saving time during open-systems approach make modular Oracle Secure Global Desktop to provide a procurement. Daily operations are also systems attractive across a broad set virtualized enterprise-class environment. simplified because one vendor can be of applications and activities — from Oracle VM Server for x86 is a fully called for support regardless of which consolidating infrastructure through supported platform for Oracle software hardware or software component has virtualization to deploying dynamic deployment and third party software caused the problem. A single support enterprise applications or cloud deployment, runs on Oracle and third ticket with Oracle means that customer IT computing. Oracle’s Sun Blade 6000 party hardware, and is free to use. teams do not have to spend their valuable time trying to isolate the right system Modular System and Sun ZFS Storage Intel and Oracle have implemented a component before calling Oracle for Appliances provide the promised joint Cloud Computing facility, the Intel support. advantages of modular architecture while Oracle Bid Support Center, which provides retaining essential flexibility for how an example of a private cloud hosted technology is deployed and managed.

46 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Additional Considerations or writes. Both Oracle’s Sun ZFS Storage The following references explain Appliances have read-optimized and write- optimization techniques that can be Storage optimized SSDs, which enable excellent implemented by using Intel SSDs and Intel-based Storage Servers support response time and throughput for compression libraries: multiple protocols such as NFS, Common demanding virtualized environments. This • http://software.intel.com/en-us/articles/ Internet File System (CIFS), Internet especially boosts VM cache performance hadoop-and-hbase-optimization-for- Small Computer System Interface (iSCSI), because of the low latency and higher read-intensive-search-applications/ InfiniBand (IB) Technology, and Fibre performance possible with SSD or flash Channel (FC). memory technology. These platforms also • http://www.slideshare.net/cloudera/ hw09-optimizing-hadoop-deployments NFS offers the utmost simplicity have clustering capabilities to provide high • http://www.cloudera.com/videos/hw10_ in attaching storage in Oracle VM availability for the storage. video_optimizing_hadoop_workloads virtualization environments over Ethernet. To address the synchronous write NFS on Intel-based Storage Servers scale performance of Oracle VM, it is strongly Scalability to many concurrent I/O threads due to recommended to have two or more write- The following guidelines are the innovative architecture design of Intel optimized flash devices per storage pool. recommended for designing and sizing the Xeon processors. This high I/O throughput To fully utilize the hybrid storage pool server pool: enables more VM stacks to perform I/O model with shorter read response time, it without sacrificing service levels. • Plan for excess resource capacity at the is recommended to use two or more read- pool level to support advanced features Architects can also choose FC or iSCSI optimized flash devices. such as Live Migration and guest VM protocols for storing and accessing both One project per Oracle VM server pool HA. Oracle VM and user data on Oracle’s Sun with dedicated storage repository is • When determining the number of ZFS Storage Appliances. While additional recommended. The shared Oracle VM nodes in a server pool, consider storage SAN infrastructure is required to use FC storage repository (/OVS) could be a file topologies and their characteristics as interface, NFS and iSCSI can run on the system or a LUN that is seen from the well as network requirements, workload existing LAN infrastructure over 1GbE or Oracle VM servers in the server pool. characteristics, and HA needs. 10GbE links. If choosing FC or iSCSI, it is • Consider that using Preferred Server recommended to use OCFS2 clustered file For storing the structured user data such VM placement policies may increase the system over the LUNs. as databases, which are accessed from the virtual machine, it is recommended number of nodes required in a server Oracle’s Sun ZFS Storage Appliances offer to match the share record size with pool. many choices in terms of RAID layout the application block size for optimal • Plan excess capacity according to to address capacity, protection, and performance (for example 8kB for OLTP business requirements for meeting performance. Mirrored or triple mirrored databases). The unstructured data and peak loads versus only a proportion of protection is the recommended RAID binary files can be stored on the shares the peak load. layout for Oracle VM storage repository with 128kB. • Memory capacity is the most critical and user data. However, depending on resource. I/O capacity and then CPU Intel® Solid-State Drives (SSDs) should be the capacity requirement and the service capacity are the second and third used to improve data intensive application level agreement, RAID Z2 (double parity priorities respectively. CPU and I/O performance by reducing I/O bottlenecks RAID) can also be deployed for storage should be balanced given that I/O and to minimize power consumption in repositories. activity is often CPU intensive. high density computing environments. • The amount of memory required for The high-performance storage capabilities Data intensive distributed applications all running VMs must never exceed can be implemented by using Intel solid should be optimized by using compression. the amount physically available on the state drives (SSDs) in some models of The compression overhead can be server(s) in the pool. However, the total the Oracle’s Sun ZFS Storage Appliances. minimized by using optimized libraries amount of memory required by all VMs Intel Enterprise SSDs enable rapid write such as Intel® Integrated Performance assigned to the pool (running + powered capabilities for fast data placement in Primitives (Intel® IPP). the storage pool. They optimize I/O rates by providing a fast buffer for reads and/

47 Oracle Optimized Solution for Enterprise Cloud Infrastructure

off) may exceed the total physical hardware, it must pass those requests to The Intel Oracle BSC environment has 1Gb memory since powered off VMs do not the back-end driver running in Dom0. The networking, however Intel ran four cables consume memory. back-end driver, in turn, receiving the I/O between each Intel architecture based • CPU over commitment is supported requests, validates them for safety and blade cabinet and the BSC network switch. (e.g. more virtual CPUs configured than isolation. Then it proxies them to the real This had the benefit of redundancy of physically present). device. When the I/O operation completes, the link and also increased the available • The over commitment ratio the back-end driver notifies the front-end bandwidth between the blades and the is dependent upon workload driver that the operation was successful network switch. In our environment this requirements. However, the and it's ready to continue. The front-end quadrupled the speed when transferring recommendation is to keep the ratio driver then reports the I/O completion to large Oracle VM image files between of virtual CPU to physical CPU at 2:1 or the guest OS kernel. servers – this reduced transfer and backup times significantly which was very less. Oracle Linux and Oracle Solaris come with useful for the largest VMs (up to 300GB • Determining the best physical server paravirtualized I/O drivers for improved each). We did not perform further analysis node size for a pool is a combination network throughput and higher disk once this benefit was achieved, so this is of factors depending on workload I/O. Moreover, Oracle VM Windows* PV described for information only. characteristics as well as the Drivers, signed by Microsoft* for the operational and budgetary issues of the Windows Logo Program, are available to data center. improve Windows guest I/O throughput. • Use identically configured server In the Oracle Optimized Solution for nodes throughout the pool to support Enterprise Cloud Infrastructure, the Sun consistent performance and feature set Blade 6000 Modular System is equipped regardless of individual server failure(s). with a redundant pair of switched In order to support the best agility NEM (Network Express Module) 24p and flexibility, it is also recommended interconnects. Each is dedicated to one that every server node in the pool be head of the Oracle’s Sun ZFS Storage identical in capacity and configuration. Appliances. This solution leverages the This approach ensures that no matter bandwidth of 10GbE while ensuring no where live migration or (re-)start single points of failure. It also offers occurs, the same performance and numerous unused 10GbE uplinks per features are uniformly available. NEM for seamless integration into the Networking datacenter. Oracle VM provides flexible networking The Oracle VM server network configuration that can be used as-is or configuration and data center topology modified to meet customers’ business should provide enterprise grade requirements. redundancy and failover. This is typically The management domain (Dom0) of the achieved by providing dual paths and Oracle VM server has direct access to switches using bonding on virtual machine the physical devices, it exports a subset servers. These settings can be configured of the devices in the systems to guest in the Oracle Optimized Solution for domains (DomU). A virtual device driver Enterprise Cloud Infrastructure. (also known as front-end driver) appears Additional details about Oracle VM to the guest operating system as a real network configurations can be found on device. The network configuration is the the third-party document called, “The same way and will look like a regular host Underground Oracle VM Manual”. with MAC address, IP address, etc. It can receive I/O requests from its kernel, but since it does not have direct access to the

48 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Processor Description

Sun Blade Server Module Features Sun Blade X6270 M2 Sun Blade X6275 M2 10GbE Specification Category Server Modules Server Modules CPU Power Power • Up to 12 cores • Up to 24 cores • Up to two six-core high-speed Intel® • Two compute nodes per blade Xeon® Processor 5600 series CPUs • Two six-core high-speed Intel® Xeon® Processor 5600 series CPUs per compute node

Main Memory • 18 DIMM slots • 12 DIMM slots per compute node • Up to 144GB total memory • Up to 96GB memory per compute node or a total of 192GB per blade

I/O • Two embedded 10/100/1000-T Ethernet • Two models: Integrated 10GbE or 1GbE networking ports for standard GbE networking • Support for high IOPS Sun Flash Modules • Total 282Gb/sec I/O bandwidth to easily handle VM networking requirements

Table 3

The following sections describe key Many server applications lend themselves it can do so through a high-speed Intel® Intel® processor technologies that enable to parallel, multi-threaded execution. QuickPath Interconnect (Intel® QPI) that implementation of virtualized multi-tenant Intel® Hyper-Threading Technology links all the processors. data centers: enables simultaneous multi-threading within each processor core, up to two Intel® Intelligent Power Technology Parallelism threads per core. Hyper-threading reduces Within a single server, Intel© Intelligent Intel-based Enterprise servers include computational latency, making optimal use Power technology minimizes power multiple processors and multiple of every clock cycle. For example, while consumption when server components computation engines to enable high- one thread is waiting for a result or event, are not fully utilized. Integrated Power performance computing. Two socket another thread is executing in that core, Gates allow individual idling cores to be servers are the typical server model to maximize the work from each clock reduced to near-zero power independent for most server applications. Four cycle. of other operating cores. Automated socket systems are used for the server Low-Power States automatically put workloads that require the higher Intel® QuickPath Technology processor and memory into the lowest performance. The number of processors Intel® QuickPath Technology is a scalable, available power states that will meet the per server can be increased if more shared memory architecture that delivers requirements of the current workload. processing power is needed for the server a high memory bandwidth to enable top Processors are enhanced with more and workload. performance for bandwidth-intensive lower CPU power states, and the memory applications. It provides high speed point- and I/O controllers have new power Intel® Turbo Boost Technology to-point connections between processors, management features. Intel® Turbo Boost Technology allows and between processors and the I/O hub. processors to deliver higher speed Each processor has its own dedicated Intel® Virtualization Technology execution on demand using available memory that it accesses directly through Intel® Virtualization Technology (Intel® VT) power to run at a higher frequency. an Integrated Memory Controller. In cases enhances virtualization performance with where a processor needs to access the new hardware assist capabilities across Intel® Hyper-Threading Technology dedicated memory of another processor, the following elements of the server:

49 Oracle Optimized Solution for Enterprise Cloud Infrastructure

• Processor: Intel VT for IA-32, Intel® 64 and storage infrastructure. Two virtualization and cloud-based use models and Intel® Architecture (Intel® VT-x) technologies comprise Intel VT-c: by helping protect servers prior to OS provides hardware assisted page-table Virtual Machine Device Queues, which launch or hypervisor launch. Intel® TXT management, allowing guest OS more accelerates throughput and reduces the complements other malware protections direct access to the hardware and load on the VMM and server processors, such as anti-virus and intrusion detection reducing compute-intensive software and PCI-SIG* SR-IOV, which delivers to help ensure that only trusted software translation from the VMM. Intel VT-x near-native throughput and provides is on the platform. VMs on trusted also includes Intel VT FlexMigration dedicated, direct connectivity between platforms are also protected, so they can and Intel VT FlexPriority, which are VMs and hardware resources. easily be migrated onto other trusted platforms or created pools of platforms capabilities for flexible workload Intel® AES New Instructions migration and performance optimization (Intel® AES-NI) Technology with trusted hypervisors. across the full range of 32-bit and 64- Intel® Advanced Encryption Standard bit operating environments. Technology enables robust encryption • Chipset: Intel VT for Directed I/O (Intel® without needing additional appliances and VT-d) helps speed data movement and increased performance overhead. This eliminates much of the performance technology improves CPU performance for overhead by giving designated virtual encryption by as much as 52 percent for machines their own dedicated I/O secure Internet transactions and allows devices, thus reducing the overhead of broader use of encryption throughout the the VMM in managing I/O traffic. data center. • Network Adapter: Intel VT for Connectivity (Intel® VT-c) enhances Intel® Trusted Execution Technology server I/O solutions by integrating (Intel® TXT) extensive hardware assists into the Intel® Trusted Execution Technology (Intel® I/O devices that are used to connect TXT) addresses a critical security need servers to the data center network for all server deployments, especially

Intel Web Resources

Intel Web Resources

Intel® Xeon® Processors http://www.intel.com/xeon

Intel® Solid-State Drives http://www.intel.com/design/flash/nand/index.htm

Intel® Integrated Performance http://software.intel.com/en-us/articles/intel-ipp/ Primitives (Intel® IPP)

Intel Optimization References http://software.intel.com/en-us/articles/hadoop-and-hbase-optimization-for-read- intensive-search-applications/ http://www.slideshare.net/cloudera/hw09-optimizing-hadoop-deployments http://www.cloudera.com/videos/hw10_video_optimizing_hadoop_workloads

Intel Cloud Builders http://www.intel.com/cloudbuilders

Intel Oracle Bid Support Center http://software.intel.com/file/18807

Table 4

50 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Oracle Web Resources For more information, visit the Web resources listed in Table 5 and the white papers listed in Table 6.

Oracle* Web Resources Web Resource Description Web Resource URL Oracle* Cloud http://www.oracle.com/goto/cloud Oracle Optimized Solutions http://www.oracle.com/us/solutions/optimized- solutions-171609.html Oracle Optimized Solution for Enterprise Cloud Infrastructure http://www.oracle.com/us/solutions/enterprise-cloud- infrastructure-405145.html Oracle Sun Blade Modular Systems http://www.oracle.com/us/products/servers-storage/servers/ blades/index.html Oracle Sun x86 systems http://www.oracle.com/us/products/servers-storage/servers/ x86/index.html Sun Networking http://www.oracle.com/us/products/servers-storage/ networking/ Oracle Solaris* http://oracle.com/solaris Oracle Linux* http://oracle.com/linux Oracle Virtualization http://oracle.com/virtualization Oracle VM Templates http://www.oracle.com/technetwork/server-storage/vm/ templates-101937.html Sun ZFS Storage Appliances http://www.oracle.com/us/products/servers-storage/storage/ unified-storage/ Oracle Enterprise Manager http://www.oracle.com/us/products/enterprise-manager/ Oracle Enterprise Manager Ops Center http://www.oracle.com/us/products/enterprise-manager/ opscenter/ Oracle Premier Support for Systems http://www.oracle.com/us/support/systems/premier/ Oracle Advanced Customer Services http://www.oracle.com/us/support/advanced-customer-services/ index.html Amazon Web Services* and Oracle Announce Certification and http://www.oracle.com/us/corporate/press/173480 Support of Oracle Applications, Middleware and Databases on Amazon Elastic Compute Cloud using Oracle VM Oracle on Amazon EC2 http://aws.amazon.com/solutions/global-solution-providers/ oracle/ Oracle VM – Creating & Maintaining a Highly Available http://www.oracle.com/us/technologies/026999.pdf Environment for Guest VM The Underground Oracle VM Manual http://itnewscast.com/underground-oracle-vm-manual Oracle VM Templates Website http://www.oracle.com/technetwork/server-storage/vm/ templates-101937.html Production Workloads in Public Clouds http://aws.amazon.com/solutions/global-solution-providers/ oracle/

Table 5

51 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Related White Papers White Paper Title Web URL Architecting Oracle VM solution with Sun ZFS http://www.oracle.com/technetwork/articles/systems-hardware- Storage Appliances and Sun Servers architecture/vm-solution-using--storage-174070.pdf Sun Blade 6000 Modular Systems from Oracle http://www.oracle.com/technetwork/articles/systems-hardware- architecture/sb6000-architecture-163892.pdf Creating and Using Oracle VM Templates: The http://www.oracle.com/us/technologies/027001.pdf Fastest Way to Deploy Any Enterprise Software Oracle VM - Creating & Maintaining a Highly http://www.oracle.com/us/products/servers-storage/servers/blades/index. Available Environment for Guest VMs html

Table 6

Follow Oracle x86 Infrastructure technology and solutions at: • Oracle systems blog: http://blogs.oracle.com/infrared/ • Oracle InfraRed on Twitter: http://twitter.com/ORCL_infrared/ • Oracle InfraRed on LinkedIn: http://www.linkedin.com/ • Oracle InfraRed on Facebook: http://www.facebook.com/pages/Oracle-Infrared/ • Oracle InfraRed on OracleMix: https://mix.oracle.com/groups/

Glossary management needs of server systems Software profile: that are used in large deployments within Intel® Data Center Manager (Intel DCM): Defines non-overlapping, logical groups of data centers, such as Internet Portal nodes that are provisioned with identical Intel DCM is an SDK from Intel that Data Centers. This includes capabilities sets of software. provides policy based tools for managing such as secure power and reset control, power in the data center. Used in temperature monitoring, event logging, Power Manager: conjunction with Intel® Intelligent Node and others. The function that is responsible for Manager and integrated with management managing the power utilization in the consoles, Intel DCM provides benefits such Hardware profile: cloud data center. as increased rack density within space, Represents a non-overlapping, logical power and cooling constraints through group of nodes that are provisioned VM: fine-grained power control, and reduced with the identical set of kernel, network Virtual Machine. capital costs by right-sizing power and configuration, disk partitioning, and device Intel® Virtualization cooling infrastructure based on actual modules. Technology (Intel® VT): power consumption trends. Provides comprehensive hardware as- Infrastructure as a Service (IaaS): Data Center Manageability sists, which boost virtualization software Infrastructure as a Service is the delivery Interface (DCMI): performance and improve application of technology infrastructure —such as response times. Intel VT reduces demands DCMI Specifications are derived from network, storage, and compute—as a placed on virtualization software so that Intelligent Platform Management service, typically through virtualization. you can consolidate more applications and Interface (IPMI) 2.0, which has been widely Users subscribe to this virtual heavier workloads per server to get bet- adopted by the computing industry for infrastructure on demand as opposed ter value from your server and software server management and system-health to purchasing servers, software, data investments. monitoring. The DCMI specifications center space, or network equipment. define a uniform set of monitoring, control Billing is typically based on the resources features, and interfaces that target the consumed. common and fundamental hardware

52 Oracle Optimized Solution for Enterprise Cloud Infrastructure

Disclaimers Intel processor numbers are not a measure of performance . Processor numbers differentiate features within each processor family, not across different processor families . See www.intel.com/ products/processor_number for details . Intel® Turbo Boost Technology: Requires a system with Intel® Turbo Boost Technology capability . Consult your PC manufacturer . Performance varies depending on hardware, software and system configuration. For more information, visit http://www.intel.com/technology/turboboost Intel® Hyper-Threading Technology: Requires an Intel® HT Technology enabled system, check with your PC manufacturer. Performance will vary depending on the specific hardware and software used . Not available on Intel® Core™ i5-750 . For more information including details on which processors support HT Technology, visit http://www.intel.com/info/hyperthreading Intel® Trusted Execution Technology: No computer system can provide absolute security under all conditions . Intel® Trusted Execution Technology (Intel® TXT) requires a computer system with Intel® Virtualization Technology, an Intel TXT-enabled processor, chipset, BIOS, Authenticated Code Modules and an Intel TXT-compatible measured launched environment (MLE) . Intel TXT also requires the system to contain a TPM v1 .s . For more information, visit http://www.intel.com/technology/security Intel® Virtualization Technology: Intel® Virtualization Technology requires a computer system with an enabled Intel® processor, BIOS, virtual machine monitor (VMM) . Functionality, performance or other benefits will vary depending on hardware and software configurations. Software applications may not be compatible with all operating systems. Consult your PC manufacturer. For more information, visit http://www.intel.com/go/virtualization Intel® AES-NI requires a computer system with an AES-NI enabled processor, as well as non-Intel software to execute the instructions in the correct sequence . AES-NI is available on select Intel® processors . For availability, consult your reseller or system manufacturer . For more information, see http://software.intel.com/en-us/articles/intel-advanced-encryption-standard- instructions-aes-ni/ INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL® PRODUCTS . NO LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROP- ERTY RIGHTS IS GRANTED BY THIS DOCUMENT . EXCEPT AS PROVIDED IN INTEL’S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS, INTEL ASSUMES NO LIABILITY WHATSOEVER, AND INTEL DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY, RELATING TO SALE AND/OR USE OF INTEL PRODUCTS INCLUDING LIABILITY OR WARRANTIES RELATING TO FITNESS FOR A PARTICULAR PURPOSE, MERCHANTABILITY, OR INFRINGEMENT OF ANY PATENT, COPYRIGHT OR OTHER INTELLECTUAL PROPERTY RIGHT . UNLESS OTHERWISE AGREED IN WRITING BY INTEL, THE INTEL PRODUCTS ARE NOT DESIGNED NOR INTENDED FOR ANY APPLICATION IN WHICH THE FAILURE OF THE INTEL PRODUCT COULD CREATE A SITUATION WHERE PERSONAL INJURY OR DEATH MAY OCCUR . Intel may make changes to specifications and product descriptions at any time, without notice. Designers must not rely on the absence or characteristics of any features or instructions marked “reserved” or “undefined.” Intel reserves these for future definition and shall have no responsibility whatsoever for conflicts or incompatibilities arising from future changes to them. The infor- mation here is subject to change without notice. Do not finalize a design with this information. The products described in this document may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized er- rata are available on request. Contact your local Intel sales office or your distributor to obtain the latest specifications and before placing your product order. Copies of documents which have an order number and are referenced in this document, or other Intel literature, may be obtained by calling 1-800-548-4725, or by visiting Intel’s Web site at www.intel.com . Copyright © 2011 Intel® Corporation . All rights reserved . Intel, the Intel logo, Xeon, Xeon inside, Intel Data Center Manager, and Intel Virtualization Technology are trademarks of Intel Corporation in the U .S . and other countries . Copyright © 2011, Oracle and/or its affiliates. All rights reserved. Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective own- ers . Oracle is committed to developing practices and products that help protect the environment . This document is provided for information purposes only and the contents hereof are subject to change without notice . This document is not warranted to be error-free, nor subject to any other warranties or conditions, whether expressed orally or implied in law, including implied warranties and conditions of merchantability or fitness for a particular purpose. We specifically disclaim any liability with respect to this document and no contractual obligations are formed either directly or indirectly by this document . This document may not be reproduced or transmitted in any form or by any means, electronic or mechanical, for any purpose, without our prior written permission . *Other names and brands may be claimed as the property of others .

Copyright 2011 Intel® Corporation SKU

53