Local Area Networks for Radiology
Samuel J. Dwyer, III, Nicholas J. Mankovich, Glendon G. Cox, and Roger A. Bauman
This article is a tutorial on local area networks (LAN) area were connected to a concentrator which was for radiology applications. LANs are being imple connected by a single communication link to the mented in radiology departments for the manage central computer. This reduced the cost of data ment of text and images. replacing the inflexible point-to-point wiring between two devices (com communication links by replacing a number of puter-to-terminal). These networks enable the shar them with a single link. Second, the computer ing of computers and computer devices. reduce was relieved of all communication processing equipment costs. and provide improved reliability. tasks by special processors called front ends. Any LAN must include items from the following four In the early 1970s, large area data networks categories: transmission medium. topology. data transmission mode. and access protocol. Media for became available for connecting multiple com local area networks are twisted pair. coaxial. and puter systems and peripheral devices. These net optical fiber cables. The topology of these networks works became known as wide area networks. The include the star. ring. bus. tree. and circuit-switch cost of communication links in these wide area ing. Data transmission modes are either analog sig networks remains a significant percentage of the nals or digital signals. Access protocol methods include the broadcast bus system and the ring sys total operating cost. tem. A performance measurement for a LAN is the In the 1980s, solid state technology advances throughput rate as a function of the number of active have provided the powerful microcomputer sys computer nodes. Standards for LANs help to ensure tems and inexpensive data communication that products purchased from multiple manufactur devices of today.l" These developments have ers will operate successfully. @) 1988 by W.B. Saunders Company. permitted movement away from large time-shar ing computers to distributed processing and KEY WORDS: Local area network (LAN). transmis resource sharing via inexpensive data networks. sion medium. topology. data transmission mode. By encoding data into small packets, it can be acess protocols. throughput. and standards. transmitted over a network and reassembled at the receiving site. In the middle 1980s, single N THE EARLY 1950s, central computers owner, small area data networks were developed. I were connected directly (point-to-point) to These small area data networks are called local terminals, printers, and other computer periph area networks (LANs).s-s eral devices. The number of peripheral devices Using either a wide area network (many and communication links increased significantly miles) or a LAN (several miles), one is able to in the early 1960s with the introduction of time access any computer or computer device on the sharing computers. Time-sharing allowed a network (assuming necessary security approv larger number of users at remote sites to satisfy als). Personal computers can act as a computer their computer needs without having to travel to or as a terminal over a network. The networks are the computer facility. As these remote devices combinations of hardware and software dedi increased in number, the cost of point-to-point cated to markedly more flexible access to and links became large. from sharing of resources."!' The possibility of Two significant developments reduced the cost electronic mail over the entire network is an of these point-to-point communication links. added benefit. First, a number of remote peripherals in a small This tutorial is intended to provide an intro duction to the basics of LANs. These LANs are From the Department of Diagnostic Radiology. Univer already in place in many small and large radiol sity ofKansas Medical Center. Kansas City; and the Depart ogy departments. They are used to link two or ment ofRadiological Sciences. UCLA School of Medicine, more personal computers to each other and to Los Angeles and the Department ofRadiology. computer peripheral devices in office automation Address reprint requests to Samuel J. Dwyer, Ill, PhD, and billing systems. The terminals and other Department of Diagnostic Radiology, The University of Kansas Medical Center, Kansas City, KS 66103. devices on radiology and hospital information © 1988 by WB. Saunders Company. systems are often connected via a LAN. LANs 0897-1889/88/0101-0004$03.00/0 also move images between devices in picture
28 Journal of Digital Imaging, Vol 1, No 1 (November), 1988: pp 28-38 LOCAL AREA NETWORKS FOR RADIOLOGY 29 archiving and communication systems. No other device. Protocols are the sets of rules and attempt will be made to provide details of those conventions governing communication on a net applications other than a few diagrams. work. These protocols minimize the collisions The focus ofthis presentation is to identify and that may occur when two or more users attempt introduce the components of a LAN and how to transfer data simultaneously. they work together. These LAN components are Historically, digital communication technol the transmission medium, the topology, the data ogy for a LAN began with the packet-switching 12 13 transmission mode, the access protocol, and the ALOHA radio network in HawaiL • The network nodes (computers and other devices). ALOHA network was established to provide Because there are many possible ways to com access for a large number of terminals connected bine components from these categories, it is to a central computing facility. Packets of data virtually impossible to simply name a LAN were broadcast on a radio channel and received without many qualifiers (exception: a commer by the central computer installation. Ethernet cial product name). Talking about a fiberoptic was subsequently developed replacing the radio LAN gives information only about the transmis channel by coaxial cable.":" Many subsequent sion medium, a very incomplete description. LAN designs have provided significant improve Teasing apart the various component parts of ments. Increased demand for LANs includes LANs will help to remove some of the mystery widespread applications for interconnecting that the many terms tend to create. large computers, workstations, and business The transmission medium of a LAN is the equipment in an office in the same building, physical medium used for the interconnection. microcomputers and digital devices in the same The electrical signaling rate on a LAN over building or between buildings, and image man various transmission media ranges from I to 50 agement networks. megabits per second (Mbps). The actual Radiology departments are already using per throughput of data transferred between two con sonal computers, file servers, printers, and LANs nected devices varies from 50 kilobits per second for a number of office automation functions. (Kbps) to 6 Mbps. This disparity between signal Figure 1 illustrates a LAN in a professional fee ing rates and throughput data rates is due to the patient-billing system which manages patient overhead necessary to operate a LAN. The topol billing, insurance claims, on-line history, and ogy of a LAN is the connection arrangement of transaction files. Furthermore, it acquires the computers and computer devices. The access patients' demographic data from the Hospital control method is the user communication proto Information Management System, transmits col for transferring data on the LAN. Every third-party payment requests, and generates connected device can communicate with every revenue reports. The computer nodes are micro-
Electronic Claims
Hospital 14----. SMS System
Fig 1. A LAN for a profes sional fee patient billing sys tem. 30 DWYER ET Al
Hospi tal Nuclear Film IMS Med. Room
286 Micro
Roentgenology
'------1AIi~be MRI
Sono
CT/Specials
Portables
Network Controller Roentgenology
Professional Fig 2. A LAN for a radiol Fee ogy information management Billing System system. computer workstation systems, a file server nodes provide for the on-line, high access rate microcomputer, a microcomputer for running storage and retrieval of the digital data. Long the management programs, and a laser printer term archiving nodes provide medium access rate (15 pages per minute) for printing patient bills, storage and retrieval of digital images. Interfaces claim forms, and reports. The LAN is a 2.5 to the department and hospital information sys Mbps token-passing network using a distributed tems are required. star topology. This professional fee billing system processes 450 patients per day, or an average of TRANSMISSION MEDIUM FOR A LAN 625 examinations per day. Commonly used transmission media for LANs Figure 2 illustrates a LAN for a radiology are twisted pair, coaxial, and optical fiber information management system that can pro cables.":" A twisted pair consists of two insu vide information about patient scheduling, flash lated wires arranged in a regular spiral pattern. card generation, film file room cards, manage The wires are copper, which provides electrical ment reports, and can track the patient through conductivity, or steel coated with copper. Twist out the department. The computer nodes are ing of the wire pairs minimizes electromagnetic microcomputer workstations, a microcomputer interference. Twisted pairs are widely used for file server, and printers. The LAN is a 2.5 Mbps analog telephone service in "local loops" that token passing token passing network with a dis connect all phones in a limited geographic area to tributed star topology which interfaces to the a central exchange. When twisted pairs are used professional fee billing system and to a hospital for digital data transmission, signaling rates of 4 information system. Mbps are attainable. Important advantages of Figure 3 illustrates a LAN for a radiology twisted pair are easy installation and low cost. A image management system. An image manage disadvantage of a twisted pair is its susceptibility ment system acquires digitally formatted imag to interference and noise because it easily couples ing data from the imaging modalities. Digital with outside electromagnetic fields. imaging data are transmitted throughout the A coaxial cable consists of two conductors; a network.P:" Interactive diagnosis display sta hollow outer cylindrical conductor surrounds a tions provide for the display and manipulation of single inner wire conductor. The inner and outer the data. Hardcopy nodes generate analog film conductors can be either solid or stranded. The or paper recordings of the image data. Archiving inner conductor is held in place by either regu- LOCAL AREA NETWORKS FOR RADIOLOGY 31
Department and ~ Hospital Information System
NETWORK
Digital Radiography DSA
Nuclear Medicine Ultrasound
CT MRI
Fig 3. A LAN for a radiol ogy image management sys tem. lady spaced insulating rings or a solid dielectric which is a solid state device that produces a material. The outer conductor is covered with a narrow light beam." Two types of photodiode jacket or shield. A single coaxial cable has a light detectors are used at the receiving end to diameter of 0.4 to 1.0 inch. A 50-12 characteristic convert the light into electrical signals, namely, a impedance coaxial cable is used for digital trans player intrinsic layer-N layer photodiode detec mission.i'r" A 75-12 characteristic impedance tor, which has a segment of intrinsic silicon coaxial cable is used for analog and digital between the P and N layers of a junction diode, signaling. These cables are used for community or an avalanche photodiode detector, which is an antenna television (CATV).23 avalanche solid state photodiode device." The An optical fiber is a thin (2 to 125 Jim), chief advantage of the optical fiber cable is that flexible medium capable of conducting light." it is not affected by electromagnetic interference Various glasses and plastics have been used to or noise. Fiber optic links are capable of provid make optical fibers. Minimal losses have been ing light transmission over 6 to 8 km without sustained using fibers of ultrapure fused silica. repeaters. They can carry signals at rates of 10 to An optical fiber cable has a cylindrical shape and 256 Mbps. Optical fibers require precision align consists of three concentric sections. The core is ment in connecting two cables, thus splicing and the innermost section and consists of one or more tapping are relatively expensive. very thin strands of fibers made of glass or The choice of a transmission medium for a plastic. The cladding is a coating surrounding LAN depends upon its applications. For text each fiber whose optical properties differ from data systems (Figs 1 and 2), twisted pair cables those of the core. The jacket is an outer layer are excellent choices. Signaling rates of 1 to 4 which protects against moisture, crushing, and Mbps provide adequate bandwidth to support other damage. Fiber optic cables are usually tens of terminals. In addition, twisted pair cables packaged in groups of two or more to make up a are inexpensive to reroute. Coaxial cables are the multi-fiber cable. choice when supporting signal rates of 10 to 50 Two types of light sources are used in fiber Mbps. Their bandwidth is capable of supporting optic systems, namely, a light-emitting diode, hundreds of terminals. They can be used for which is a solid state device that emits light when digital, analog, or mixed signaling formats. How a current is applied, or an injection laser diode, ever, coaxial cables are expensive and difficult to 32 DWYER ET AL
Computer install. A number of commercial LAN systems Node now use coaxial cables. Optical fiber cables pro vide the highest bandwidth for signaling, ranging from 10 to 256 Mbps. The increasing use of optical fiber cables into LANs makes this medium the future choice for radiology image management applications. Computer Computer Nod. Nod. TOPOLOGY OF A LAN The topology of a LAN is the arrangement of the computers and computer devices and the 3 interconnections among them. ,5,6 The impor Computer Computer Node Node tance of the topology of a LAN is illustrated by A determining the number of links needed to inter Computer connect N devices fully. There will be N(N-l)j2 Node links required to interconnect each device to all of the other devices. For example, if we were to interconnect ten devices (N = 10) so that each Computer device was linked to all the other devices, then a Node maximum of 45 links would be required. Hence, the cost of interconnection in a LAN depends significantly upon the choice of topology of the network. The topologies used in a LAN are the following: star, ring, bus (baseband, broadband), tree, and circuit switched. These arrangements connect all nodes with a varying number of links Computer Computer but in each case, it is less than the maximum Node Node number of links. A star LAN topology is illustrated by Fig 4A. B Computer Communication to any connected device is con Node trolled by the hub. Device "w" wishing to trans computer Computer Computer mit to device "X" sends such a request to the Node Node Nod. hub. The hub inserts this request into a service o queue. When the queue services this request, devices X and W then transmit data according to their communication protocols. No interference among other users occurs since the hub estab lishes the data paths and blocks other paths until C Computer Computer the X-to-W communication is complete. The Node Node advantages are that addressing each computer Fig 4. (A) A LAN arranged in a star topology. (BI A node is easy, all computer nodes are centrally LAN arranged in a ring topology. (CI A LAN arranged in a controlled by the hub, and it is easy to detect bus topology. faults. The disadvantages include vulnerability to hub failures, data throughput rates that are plished by device W inserting its data onto the limited by hub speed, and the high cost of the ring. The ring transmits this data from ring centralized hub. Future LAN topology will move access point to ring access point. Each ring access towards the star topology as the link bandwidths point reads the destination address of the data increase, hubs decrease in price, and program packet. Ifthe address is correct, then the data are mable link data rates are implemented. read into the node and removed from the ring. If A ring topology is illustrated by Fig 4B. Com the address is not correct, then the ring access munication from device W to device X is accom- point transmits the data packet to the next ring LOCAL AREA NElWORKS FOR RADIOLOGY 33 access point. The advantages are: data transmis LAN described is a LAN based upon high-speed sion capacity is shared uniformly among all circuit-switching devices similar in operation to users; there is no dependence on a central device; those of the telephone switched systems.t-" A high data transmission rates are possible; access circuit-switched LAN provides interconnections by the user is guaranteed, even when the ring is among devices by switching the sending device to heavily loaded; and it is inexpensive to imple the receiving device. These switching devices are ment. Disadvantages include: breakage of any configured in a star topology using twisted pair link halts all data transmission; cable installation wires to connect end points to the switch. Data and routing can be difficult; modification or rates to individual stations are low (64 kbps), but rerouting of the ring is difficult; each repeater at the bandwidth is guaranteed once a connection is every computer node or device imposes some made. One form of circuit-switched LAN is the signal delay; and a ring network monitoring digital private branch exchange. The availability device is usually needed to maintain smooth of a high-speed circuit-switching device would operation of the LAN. provide LAN capabilities without complex Figure 4C illustrates a LAN using bus topol medium access protocols. ogy. Communication on a bus topology depends upon all nodes listening to all signals on the bus. DATA TRANSMISSION MODES When device W wants to transmit to device X, it Data are defined as those entities that convey begins transmitting its data. Device W listens to meaning, content, or interpretation of the data." the bus during its own transmission. If it hears Analog data take on continuous values over some itself transmitting accurately, then it assumes interval, eg, voice or video. Digital data take on the message was correctly received (assuming discrete values over some interval, ego integers or device X is not asleep). Ifthe signal it hears is not text. A channel is a medium such as a wire, a the same as the one being transmitted, then coaxial cable, or an optical fiber. The bandwidth device W assumes a collision has occurred, ie, of a channel is the range of frequencies that it someone else is also attempting to transmit on can transmit with reasonable fidelity. For exam the same bus. The advantages of bus topology ple, a telephone channel bandwidth is 300 to are: it is easy to add additional nodes; the bus is 3400 Hz, while voice data are in the range of 20 passive, thereby improving reliability; coaxial Hz to 20,000 Hz. The limitations of channel cable is used (either 50 Q or 75 Q); and this bandwidth determine the signaling rates and configuration is widely available. Disadvantages data throughput rates. include: the access communications protocol to The distortion of the input waveform to any transmit data is complex; network interface units communication link is caused by amplitude require buffer storage in order to achieve high frequency distortion, which is the result of non data throughput rates using small packet sizes; uniform attenuation of the frequency spectrum and data throughput declines with heavy loads. of the transmitted signal, or phase-frequency A LAN using a tree topology consists of a very distortion, which is the most limiting impairment high data rate baseband bus (256 Mbps), called to data transmission and is caused by a nonlinear a backbone, with multiple drop-off baseband phase shift of the frequency spectrum of the busses (10 to 50 Mbps). This type of LAN transmitted signal." Distortion is also caused by topology is suitable for department-wide image nonlinearities in data channels and low signal 2 22 management networks, and it is capable of com to-noise ratios. 1. The small distances covered bining fiber optics, coaxial cables, and twisted by a LAN result in small amplitude-frequency wire pairs. Because of its costly backbone link, a distortion, small phase-frequency distortion, and tree topology LAN is expensive to implement. large bandwidths. Small distances and careful Communication protocols are complex and attention to shielding result in high signal-to require a communication controller. However, a noise (SIN) ratios. The relationship between tree topology is the likely result of the growth and bandwidth and SIN is given by the Shannon integration of many small LANs into a single Channel Capacity Theorem," ie, C = B log, network. (1 + SIN) where C is the maximum channel An alternative to the packet-switched type capacity in bits per second, B is the bandwidth of 34 DWYER ET AL the channel (hertz), and (N) the mean square tages are that only a single channel is available, noise of the channel, which is additive and white only limited distances may be covered, and gaussian distributed. capacity for high speed transmission is limited. Data are encoded for transmission on a chan Broadband systems modulate information on nel into either analog signals or digital signals. an analog carrier wave. When applied to a LAN, The basis for analog signaling is a continuous the term broadband means the use of FDM and constant-frequency signal known as the carrier coaxial cable on which analog signaling is used. signal. Digital data are encoded by modulating The advantages of broadband include: high data one of the three characteristics of the carrier transmission capacity, large area coverage, the signal, ie, amplitude (amplitude-shift keying); use of mature CATV technology, and the ability frequency (frequency-shift keying); or phase to mix data traffic types such as voice, FAX, (phase-shift keying). Analog signaling is used for video images, or data. The disadvantages lower signaling rates such as 1,200 bps or 19.5 include: modulation equipment is expensive, kbps. The basis for digital signals is the use of installation and maintenance is costly, and the two different voltage levels for the two binary propagation delay is twice that of baseband digits. One voltage level is used to represent a systems. Another example of FDM is the analog "zero," while the other voltage level represents a voice telephone system. A bandwidth of 4 kHz is "one." The Manchester Code is most often used allocated for each voice channel. 5 6 in a baseband LAN. • This code is a self clocking code in which a signal transition occurs ACCESS CONTROL METHODS during each data bit to enable synchronization of The responsibility for allocating the LAN the hardware. The bit is a zero if a low-high capacity is usually distributed among all the transition occurs within the bit time interval and nodes on the network.":" No single node is used a one if a high-low transition occurs. as an access controller. The common LAN access It is desirable to use the LAN medium effi control methods are the broadcast-bus system ciently by having it carry multiple signals simul and the ring system.Y taneously. This is called multiplexing. The three A broadcast system is one in which a message, methods of multiplexing are: (l) time division or packet containing part of a message, is sent multiplexing (TDM), also called baseband sig into the LAN and can be heard all over it. The naling, in which each device is allocated a unique signal rapidly dies away once the transmission time slot; (2) statistical TDM in which time slots has ceased. In a broadcast system, all the users are allocated to devices when needed; and (3) are capable of hearing everything transmitted to FDM, also called broadband signaling, in which them, to another user, or to everybody. When two each device is allocated a unique band or range of packets of data are broadcast at the same time, frequencies. they are said to have collided, and each packet is An example of the TDM is the pulse code damaged, ie, data are corrupted. A node wishing modulated TI carrier system. In the TI carrier to transmit listens to the medium (carrier sense) system, data are taken at each source and digi and performs the following: (I) if the medium is tized to seven bits, then an eighth bit is added for idle, then it starts transmitting data; (2) if the signaling control. Twenty-four sources are multi medium is busy, then it listens until the channel plexed so that each frame is 192 bits wide is sensed to be idle (carrier sense) and subse (8 x 24). Data are taken at each source 8,000 quently starts transmitting; and (3) if a collision times per second. The data rate of a TI carrier (two nodes start at the same time) is detected, it system is 1.544 Mbps ([192 + I frame waits a random amount of time and then repeats bit] x 8,000). steps I through 3. The use of steps I through 3 is Baseband techniques have been widely called a carrier sense multiple access (CSMA) adopted by LAN designers because the signal bus. CSMA with collision detection (CSMAj can be transmitted at high rates and no modula CD) uses a slight modification to step 3. If a tion technique is required. The advantages are collision is detected during transmission, it that it is inexpensive to implement, the technol immediately ceases transmitting the frame of ogy is simple, and it is easy to install. Disadvan- data and instead transmits a brief jamming LOCAL AREA NETWORKS FOR RADIOLOGY 35 signal to ensure that all stations know that there to ensure that promised capabilities are realiz has been a collision. The amount of time it takes able. Three measures often used are: (l) the to detect a collision is twice the propagation throughput of the LAN, denoted by "S," ie, the delay at the network. After transmitting the data transfer rate between nodes, often called the jamming signal, it waits a random amount of carried load; (2) the utilization of the LAN time and then transmits again using the CSMAj medium, denoted by "U," ie, the fraction of total CD access control protocol (ie, Ethernet). capacity being used; and (3) the delay time, Ring access techniques are accomplished denoted by "D," between initiating transmission through a repeater placed at each node. Repeater and the completion of successful transmission. X receives the data packet passed to it by The parameter S is often normalized and repeater W. Repeater X makes the packet avail expressed as a fraction of the capacity of the able to the devices attached to node X. It then LAN. allows all devices at node X to transmit a data The performance of a LAN is the result of the packet by transmitting the packet to the next interaction of the four categories discussed and repeater, "Y." A special ring access technique, other major factors including the number and called token passing, passes a specific bit type of nodes, their activity, and the amount and sequence, called a token, from node to node character of data on the network. Whether the around the LAN. A node can transmit only when messages are short (small packets) or long (large a token is passed to it. Only when its communica packets), they can have major impact on LAN tion is complete does the node pass the token performance which will vary with different LAN again. The advantages of the token ring access configurations. Small packet sizes are common method are: no packet collisions are possible; in office automation. Radiology images require utilization rate of the LAN is uniform; and large packet sizes. The remainder of this section hardware implementation is simple. Token ring on performance is a more technical discussion. access methods may also be used on broadcast The use of queueing analysis provides the sto bus topology by allowing a node to transmit only chastic tools needed to analyze the performance when it possesses the token. of a proposed LAN.32 However, a simple tech The choice of an access method depends upon nique is often used to estimate the upper and the LAN application. For data packets consist lower performance bounds of the LAN proto ing of text (Figs I and 2), it is best to use a star cols." The assumption is to ignore the medium topology and a token ring access method. This access control protocol and to know or estimate provides good throughput at a reasonable cost. the following mean times: (1) TJ, the mean time For LAN applications where the data records are that a node is idle between transmission long, eg, image management (Fig 3), the best attempts; (2) T M, the mean time needed to choice depends upon the extent of the depart transmit a message once access to the medium is ment-wide coverage. For an image management acquired; (3) T D, the mean delay from the time a LAN in some sections of a department, the node has a packet to transmit until transmission broadcast bus with CSMAjCD or token access is completed; and (4) S, the mean total through might be best. A department-wide LAN image put on the network, ie, messages per unit time. management network will probably require a Assume that there are N active nodes on the tree topology with fiber optic cables. It is likely network, each identical with the same transmis that departments will slowly enlarge LAN image sion requirements. An upper bound is found for management systems by implementing the net the total throughput, 8, by assuming each node is work in the department section by section. It will in either an "idle" state (T1) or in a "transmis be necessary to bridge together the image man sion" state (TM)' The mean throughput for each agement of LAN and the patient data LAN. station is S = Ij(T1 + T M ). If no queues occur, ie, no node is waiting for another node, then the PREDICTING LAN PERFORMANCE maximum possible throughput will be 8 = N j
The performance of LAN protocols has been (T 1 + T M ). Figure 5 illustrates the upper bound extensively studied.F' 30-32 Careful analysis of the on the mean throughput, 8, as a function of N. performance of a LAN is difficult but necessary However, as more nodes become active, we soon 36 DWYER ET AL
T.:pperBound of computers, computer devices, and distributed S, Throughput networks has forced the acceptance of standards. A standard is a published set of rules or require ments concerning the terms, components, mate ~ LowerBound rials, performance, and operations in describing products, services, or systems." A number of standards have helped establish the LAN indus try. Ns N,numberofnodes The International Organization for Standard Fig 5. The upper and lower bound on a LAN mean ization (ISO) was founded in 1946 and has throughput as a function of the number of active nodes. N. issued over 5,000 standards on a broad range of subjects. Technical Committee 97, the informa approach the saturation value of N = Ns. At this value of N, we no longer have any idle time tion processing system of ISO, published in 1984 the Open Systems Interconnection (OSI) model remaining. The break point occurs when Ns/ of a computer communications architecture. The (T 1 + T M) = l/TMor Ns = (T 1 + TM)/(TM). This simple scheme permits us to size a LAN OSI model is accepted and is followed by all protocol for upper bounds on the throughput. LAN manufacturers. The structured set of pro A lower bound on mean throughput as a tocols of the OSI model are layered as shown in function of N can be established by assuming Fig 6. The purpose of the OSI model is to devise that all nodes have a message to transmit simul protocols to perform the functions of each layer. taneously, then the lower bound for the mean The communication protocols are partitioned throughput, ~, would be given by ~ ~ N/ into a vertical set of layers, and each layer provides services to the next higher layer. The (T 1 + NTM). This is so because we are assuming that each node desiring to transmit a message lowest layer, the physical layer, specifies the will have to wait until all other nodes have electrical and procedural standards for accessing transmitted. The lower bound, ~, is shown in Fig the network. The data link layer provides stan 5. The actual value of the mean throughput, S, dards for the reliable transfer of user data over will reside somewhere between the upper and the data link. The network layer determines the lower bounds. most appropriate network configuration for the The above method provides a simple model for function being serviced, ego dial up line or packet estimating the mean throughput of a LAN. switching. The transport layer controls end-to Improved estimates can be calculated by replac end functions such as routing, error recovery, and ing TMwith Tsy., where T sys is the time required to transmit a message plus the overhead time per I Outgoing Incoming + frame frame Computer packet. More accurate estimates can also be Node "X" t attained by considering the medium access pro tocol. This type of analysis has been carried out for the protocols of CSMA/CD, token bus, and 27 32 Presentation Presentation token ring. • Layer Layer
Session Session LAN STANDARDS Layer Layer
Transport Transport The communication industry long ago learned Layer Layer that standards are required to govern the physi Network Network cal, electrical, and communication protocols for Layer Layer Data Link Data Link transmitting data." This lesson was forcefully Layer Layer taught since marketed communication equip Physical Physical Layer ment had to interface to that of other manufac Layer ~ J turers. The computer industry was slower to Communication Media accept the need for standards since each manu facturer wanted to provide all needed equipment. Fig 6. The seven layers of the Open Systems Intercon However, the growing number of manufacturers nection (OSII model. LOCAL AREA NETWORKS FOR RADIOLOGY 37 message segmentation. A session layer ensures and integrated services digital networks. The the availability of resources such as file storage American National Standards Institute is a and processor time. The presentation layer clearing house for standards in the United includes encoding, formatting, interruption and States. termination. The application layer commu nicates directly with the user's program or ser DISCUSSION vices. A LAN uses the lowest two layers plus The declining costs and widespread applica portions of the third layer of the OSI model. tions of LANs make them an ideal extension of In 1980, the Institute of Electrical and Elec communication technology for use in radiology tronics Engineers (IEEE) established the IEEE departments. Radiology departments now use 802 Local Area Network Committee to stan portions of LAN technology in computer, office, dardize the means of connecting digital com and information management systems. The puter equipment and peripherals to networks. In implementation of a LAN, using personal com 1983, the IEEE standards committee board puters and file servers, offers an attractive alter approved standard 802.3, the CSMAjCD proto native to a system consisting of a large computer colon a bus topology and standard 802.4, the with dedicated terminals. LANs are being stud token passing protocol on a bus topology. The ied and developed for the department-wide man 802.2 standard is a logical link control. The 802.5 agement of patient images and consultation data. standard, which details token passing protocol on The increasing demand for wide-spread applica a ring topology, has been approved. The 802.1 tions of LANs ensures that radiology will benefit document describes all the 802 standards and from continued cost reductions and improved their relationship to the ISO Open Systems capabilities. Interconnection Reference Model. All the IEEE The technology of LANs is rapidly changing. 802 standards are supported by the ISO. Each month brings announcements of technol In 1985, a committee of the American College ogy improvements and new manufacturers. The of Radiology (ACR) and the National Electrical half-life of LAN technology is shorter than that Manufacturers Association (NEMA) published of imaging systems in the department. If a the ACR-NEMA Standard 300-1985, "Digital radiology department selects a LAN manufac Imaging and Communications.Y" This standard turer, what assurances are there that the manu specifies the hardware interface, software com facturer will be in business the following year? mands, and data formats for communication The protection afforded by the OSI interconnec across an interface between an imaging device tions standard ensures that any manufacturers' and a network interface unit. Additional stan equipment can be substituted. The best protec dards are under development by the ACR tion is to require department staff members to NEMA committee, including digital data for become knowledgeable in LAN technology and mats, data compression, and LANs for image its uses. management. Software remains the most expensive and dif Other standard-making organizations in ficult aspect of using and implementing a LAN. fluence the LAN standards. The International Integrating software packages into an informa Telegraph and Telephone Consultative Commit tion or image management LAN is difficult to tee (CCITT) provides standards dealing with the accomplish but the power and advantages telegraph and telephone. CCITT standards offered by LANs will undoubtedly lead to their include data communications, telematic services, increased implementation in radiology.
REFERENCES 1. Martin J: Telecommunications and the Computer (ed 4. Stallings W: Tutorial: Computer Communications: 3). Englewood Cliffs. NJ. Prentice-Hall. 1976 Architectures. Protocols and Standards. Silver Springs. MD. 2. Martin J: Computer Networks and Distribution Pro Institute of Electrical and Electronics Engineers. Computer cessing. Englewood Cliffs. NJ. Prentice-Hall. 1981 Society Press. 1985 3. Tanenbaum AS: Computer Networks. Englewood 5. Gee KCE: Introduction to Local Area Computer Net Cliffs. NJ. Prentice-Hall. 1981 works. New York. Wiley. 1983 38 DWYER ET Al
6. Stallings W: Local Networks (ed 2). New York, Mac 20. Baker DG: Local-Area Networks With Fiber-Optic millan, 1987 Applications. Englewood Cliffs, NJ, Prentice-Hall, 1986 7. Freeman HA, Thurber KJ: Local Network Equipment. 21. Bennett WR, Davey JR: Data Transmission. New Silver Springs, MD, Institute of Electrical and Electronics York, McGraw-Hili, 1965 Engineers, Computer Society Press, 1985 22. Lathi BP: Modern Digital and Analog Communica 8. Stallings W: Tutorial: Local Network Technology, (ed tion Systems. New York, Holt, Rinehart, and Winston, 1983 2). Silver Springs, MD, Institute of Electrical and Electron 23. Alisouskas VF, Tomasi W: Digital and Data Commu ics Engineers, Computer Society Press, 1985 nications. Englewood Cliffs, NJ, Prentice-Hall, 1985 9. Kiimmerle K, Tobagi FA, Limb JO, (eds): Advances in 24. Li T: Structures, parameters, and transmission prop Local Area Networks. New York, Institute of Electrical and erties of optical fibers. Proceedings Institute of Electrical and Electronics Engineers Press, 1987 Electronics Engineers 68:1175-1180,1980 10. Champine GA: Distributed Computer Systems 25. Bergh AA, Copeland JA: Optical sources for fiber Impact on Management, Design, and Analysis. Amsterdam, transmission systems. Proceedings Institute of Electrical and North-Holland,1980 Electronics Engineers 68:1240-1247, 1980 26. Smith RG: Photodetectors for fiber transmission sys II. Gupta A, (ed): Advanced Microprocessors, II. New tems. Proceedings Institute of Electrical and Electronics York, Institute of Electrical and Electronics Engineers Press, Engineers 68:1247-1253, 1980 1987 27. Hammond JL, O'Reilly PJD: Performance Analysis 12. Taub H, Schilling DL: Principles of Communication of Local Computer Networks. Reading, MA, Addison Systems (ed 2). New York, McGraw-Hili, 1986 Wesley, 1986 13. Kurosa JF, Schwartz M, Yemini Y: Multiple-access 28. Abramson N: Information Theory and Coding. New protocols and time-constrained communication. Association York, McGraw-Hill, 1963 of Computer Machinery (ACM) Computing Surveys 16:43 29. Green PE Jr (ed): Computer Network Architectures 70,1984 and Protocols. New York, Plenum, 1982 14. Clark DD, Pogran KT, Roed DP: An introduction to 30. Folts HC, des Jardins R (eds): Special Issue: Open local area networks. Proceedings Institute of Electrical and systems interconnection (OSI)-Standard architecture and Electronics Engineers 66: 1497-1517, 1978 protocols. Proceedings Institute of Electrical and Electronics 15. Metcalfe RM, Boggs DR: Ethernet: Distributed Engineers 71:12, 1983 packet switching for local computer networks. Commun 31. Graham GS (ed): Special Issue: Queueing network ACM 19:395-403, 1976 models of computer system performance. ACM Computing 16. Shoch JF, Dalal YK, Redell DD, et al: Evolution of Surveys 10:3, 1978 the ethernet local computer network. Computer 5:10-27, 32. Stuck BW, Arthurs E: A computer and communica 1982 tions network performance analysis primer. Englewood 17. Thornton JE, Christensen GS: Hyperchannel network Cliffs, NJ, Prentice-Hall, 1985 links. Computer 16:50-54, 1983 33. Stallings W: Handbook of Computer-Communica 18. Tropper C: Local Computer Network Technologies. tions Standards (vol I). New York, Macmillan, 1987 San Diego, Academic, 1981 34. Wang Y, Best DE, Hoffman JE, et al: ACR-NEMA 19. Freeman RL: Telecommunication Transmission digital imaging and communications standards: Minimum Handbook, (ed 2). New York, Wiley, 1981 requirements. Radiology 166:529-532, 1988