Fiber Distributed Data Interface Overview By William R. Hawe, Richard Graham and Peter C. Hayden Abstract As the use of local area After exploring various networks (LANs) continues alternatives to second- to grow at an exponential generation local area rate, many large networks networks (LANs), with Ethernet backbones Digital selected the are reaching their usable fiber distributed data capacity. In addition, interface (FDDI) system. the explosion in the FDDI implements the use of high-performance International Standards workstations is placing Organization (ISO) physical increasing demands on layer and the media access network performance as control sublayer of the larger volumes of data pass data link layer. This from station to station. system is based on a 100- Several years ago, Digital megabit-per-second fiber- recognized this growth optic ring network and uses trend and began to plan and a timed-token protocol to develop a second-generation coordinate station access LAN that would follow to the network. Digital Ethernet and provide an has developed the FDDI base evolutionary path to higher technology, including very performance. The selection large-scale integration of FDDI as the second- (VLSI) chips and software. generation LAN was made These chips, licensed to with great deliberation. Advanced Micro Devices and This paper explores the Motorola, Inc., provide criteria for that choice high-quality alternatives and the history of the FDDI in the market and foster system to the present. The cost reduction. Digital's theory of FDDI operation, implementation of FDDI, the development of the including backbones in FDDI technology's role in extended LANs, as well as Digital's networks, and the high-speed interconnection resulting products are also of workstations, servers, presented and discusssed. and central computers, makes available a complete range of system products. Introduction Digital Technical Journal Vol. 3 No. 2 Spring 1991 1 Fiber Distributed Data Interface Overview Selection of FDDI solution will not integrate Many of the same criteria well in the computer originally used to select interconnect environment Ethernet were again used and will not be cost to evaluate the application effective for wide-scale environment for the new use. LAN. The need to consider FDDI, with its tenfold migration from the popular increase in speed, provides LANs currently in use significant differentiation presented the only new from Ethernet/802.3 and concern. Paramount among current token ring and bus the reasons for selecting technology to justify the the FDDI technology were new investment. Examination its tenfold increase in of the clocking, buffering, bandwidth over Ethernet, and state machine needs of its consistency with other the media access control IEEE 802 LANs, and the (MAC) sublayer of the data standardization effort link layer also showed that already begun in the the FDDI technology could American National Standards be implemented in several Institute (ANSI). VLSI chip components. It is important when Further, as silicon developing a new LAN technology improves, cost technology to be sure reduction is possible, the differentiation from enhancing the longevity of current capabilities is the FDDI LAN technology. sufficient to warrant the necessary investment. Moreover, a new LAN is a significant investment for a customer and should offer a large increase in capabilities such as speed and throughput. Without this increase, the technology will have a short life span (a few years) and technology such as parallel use of existing LANs to double capacity will be a realistic alternative to a wholesale replacement of the LAN. However, it is important not to take such a large technological step that exotic and complex implementation constraints become necessary. A LAN that does not lend itself to a very large-scale integration (VLSI) logic 2 Digital Technical Journal Vol. 3 No. 2 Spring 1991 Fiber Distributed Data Interface Overview Migration is another between LANs. Ethernet, important factor in the the forerunner of IEEE selection of a new LAN. 802.3, does not use the Many devices exist with IEEE 802.2 formatted LLC embedded LAN interfaces and, therefore, migration that will never be directly of those frames is more connected to any other challenging. LAN. Still other devices Lastly, the media selected will not benefit from the for the new LAN has to be added capabilities of a consistent with current and higher speed interconnect. projected future needs. Examples of such devices Even for slower speed include a processor or LANs, fiber-optic media workstation too slow to is gaining in popularity send or to receive data any because of its superior faster from the network, qualities in spanning an output-limited print greater distance, its noise server, or communications immunity, and its declining servers with other, more user cost. constraining, I/O ports. It would never prove cost The FDDI technology meets effective to upgrade these the necessary selection devices to a new higher criteria as an emerging speed LAN interface but, American National Standards as a group, they would need Institute (ANSI) standard to obtain unconstrained (no using fiber and allowing bottleneck) connectivity other media in place of to the services of the new fiber while providing a LAN for smooth migration tenfold increase in speed. allowing protection of the Migration of some devices investment in devices and could be affected directly LANs already in place. by changing controllers, Standards are important and bridging between for networks as a way LANs could allow smooth to ensure consistent migration of all existing interface compatibility devices. and interoperability for communications services. FDDI History As the IEEE 802.1d standard Both the ANSI FDDI readily demonstrates by standards and the industry- attempting to interconnect wide implementations dissimilar LANs at the MAC of these standards sublayer, some standards have evolved slowly. are more compatible than A variety of factors others. A common logical have contributed to this link control (LLC) format course of development. The or the format within the FDDI ring was originally MAC frame allows a smooth invented at Sperry and migration between LANs Burroughs Corporation. The by allowing a transparent ring was to be used as a bridge to provide protocol- machine room interconnect independent translation between processors and Digital Technical Journal Vol. 3 No. 2 Spring 1991 3 Fiber Distributed Data Interface Overview storage systems, much like In 1982 the FDDI technology the Computer Interconnect was brought to the components are used in attention of the ANSI Digital's VAXcluster X3T9 committee, which systems.[1] The timed- develops standards for token, media access control I/O interconnects and protocol itself was first channels. Since FDDI was publicized in 1982 by intended to be used as a Bob Grow while he was at machine room interconnect, Burroughs.[2] this committee was the As a machine room appropriate arena for interconnect between study. Over time, however, processors and storage as the need for a 100- systems, the initial ANSI megabit-per-second standard requirements LAN emerged, the FDDI on the FDDI technology technology evolved into were quite different a local area network. Some from today's needs. In classic standards territory particular, as a machine conflicts developed between room network, the number of IEEE 802, the group that stations was assumed to be defines all the LAN relatively small compared standards, and this ANSI to a LAN and unstructured committee. cabling was to be used. While FDDI was evolving Since most machines from a machine room were always running, interconnect into a fault recovery could be general-purpose LAN, the accomplished by having a requirements changed. For a dual ring with failover machine room interconnect, to the secondary ring. some management operation Thus, a failed station or to install and initialize cable could be isolated the network might without partitioning the reasonably be allowed. For ring. The FDDI technology example, the manager might retains this property set the values of various today. However, that basic parameters to control the operation capability is operation and performance insufficient in a LAN of the interconnect environment with structured network. However, in cabling requirements and a a general-purpose LAN, large number of stations, manager involvement any number of which might is unacceptable. For be unplugged or turned simplicity, robustness, off by users. Therefore, and ease of management, we have expanded the the industry widely definition of the FDDI accepts that LANs must technology to include such autoconfigure, also products as concentrators called "plug-and-play." and adapters. Inevitably, FDDI was required to exhibit the attributes of a true local area network. 4 Digital Technical Journal Vol. 3 No. 2 Spring 1991 Fiber Distributed Data Interface Overview Since the FDDI technology and standards were already in development between stations.[4] when this evolution of The control information requirements occurred, exchanged varies with the the ANSI committee made physical layer protocol an attempt to accommodate type, which is either PHY- the following two views A, PHY-B, PHY-M, or PHY-S. of the network: first, The MAC sublayer specifies the network should be the protocols for logical completely configurable ring formation and control, with almost every parameter for the transmission and and policy controlled by management; and second, reception of packets at the network should be a station, and for the a local area network repetition and stripping of with the corresponding packets on the ring.[5,6] attributes of simplicity Station management (SMT) and autoconfiguration. provides n-layer management Incorporating both and a local management models into the ANSI interface to the PMD, PHY, FDDI standards made the and MAC layers.[7] Together standards complex and was these components support one factor contributing to an IEEE 802.2-compatible the eight-year-long time logical link control period to completion. capable of supporting client protocols such as Theory of Operation the Digital networking (DECnet) protocol, open FDDI stations are composed systems interconnection of the basic elements (OSI), local area transport defined by the FDDI (LAT), and the transmission standards. The physical control protocol/internet medium dependent (PMD) protocol (TCP/IP). layer specifies the Stations utilizing the fiber-optic interface FDDI components can take and data driver and several forms such as receiver operation for FDDI single attachment stations stations.[3] The physical (SASs), dual attachment layer protocol (PHY) stations (DASs), and dual specifies the encoding and attachment concentrators framing of data and control (DACs). An architectural information exchanged model is shown in Figure 1. Digital Technical Journal Vol. 3 No. 2 Spring 1991 5 Fiber Distributed Data Interface Overview Configurations of FDDI the additional PHY-Ms in Components the concentrator. A single attachment An FDDI LAN is formed by station is the simplest joining multiple stations configuration and consists to form a logical ring of the fundamental FDDI topology. The logical components arranged as ring can take two physical shown in Figure 2. There is forms, a dual trunk ring a single incoming data path and a tree ring. and a single outgoing data As depicted in the upper path with a MAC in between. portion of Figure 5, Dual attachment stations, the dual trunk ring is as illustrated in Figure formed by connecting dual 3, include a second attachment stations and physical layer and provide concentrators to form connections to a secondary a LAN. This portion of ring for use in the event the LAN consists of two of breakage on the primary data paths in opposite ring. Under fault-free directions, called the operating conditions primary and secondary represented by the THRU rings. Under normal STATE area of Figure 3, the operation, data flows primary data path enters on the primary ring from through PHY-A, travels station to station. In through the MAC, and the event of a cable or exits through PHY-B. The a station breakage, the secondary data path enters stations adjacent to the through PHY-B and exits fault join the primary and directly through PHY-A. If secondary rings and then a discontinuity is detected use the secondary path to in the primary data path, reestablish a logical ring. either within the station A tree ring can be formed or on one of the PHYs, the by connecting stations or station wraps the two data concentrators to the PHY- paths, thereby providing an Ms of a concentrator as alternate route through the shown in Figure 5. In this secondary data path. This formation, the primary data situation is shown in the path descends down each WRAP A STATE area of Figure branch of the tree passing 3. through each station in A dual attachment the tree, until it finally concentrator builds on the reemerges into the dual dual attachment station trunk ring. by adding additional master PHYs (PHY-M) in the primary data path as shown in Figure 4. Single attachment stations can then be included in the ring by connecting them to 6 Digital Technical Journal Vol. 3 No. 2 Spring 1991 Fiber Distributed Data Interface Overview Media Access Control To transmit data packets, Sublayer Operation the MAC first waits for the token to arrive, holds As mentioned previously, it, and, then, transmits the MAC sublayer provides the packets, reissuing a the protocols for logical new token at the end of ring formation and data the transmitted packet packet operations. To stream. The time allowance initialize the ring, all a station has to transmit MACs first enter the claim packets after receiving a process to determine which token is equal to the token MAC will generate the rotation time established token and to establish by the claim process. the token rotation time Packets received by a for the ring. Each station MAC are either repeated continuously transmits for reception by the next claim frames that contain station or stripped from the station's requested the ring. In addition, token rotation time. When a a MAC may store a copy MAC receives claim frames of a packet for use with times shorter than by the station. After its own, or equal to its transmitting a frame, a own but from a station MAC is responsible for with a numerically larger stripping that frame from address, it yields, stops the ring after the frame sending claims, and repeats makes exactly one traversal the claims received from of the ring. Frames left its neighboring station. unstripped are considered Eventually the station with no-owner frames and can the shortest or "winning" circulate the ring forever. time will receive its own This condition floods claim. This station then the station to which the generates the token and the frame is addressed and ring enters the operational is thus detrimental to state. ring performance. A MAC If the claim process typically strips frames does not complete by comparing the source within approximately 100 address in the frame with milliseconds, the MACs the MAC's own address. The in the ring perform a MAC strips any frame it has beacon process to confirm previously sent but repeats continuity of the ring. the frame, otherwise. Special beacon frames are FDDI Ring Formation transmitted continuously by all MACs until a beacon An FDDI ring is formed in is received, at which point several stages, beginning it stops transmitting. This with the successful process continues until one establishment of point- MAC transmits and receives to-point links between all its own beacon, indicating adjacent PHYs. These link ring continuity. connections are made by the connection management Digital Technical Journal Vol. 3 No. 2 Spring 1991 7 Fiber Distributed Data Interface Overview protocol (CMT).[7] This essential to many LAN protocol defines control applications such as disk signal exchanges to I/O, interactive graphics, synchronize the two ends and remote procedure calls. of the link, to exchange Higher settings of the information about the PHY requested token rotation type (i.e., A, B, M, S) and time result in very large link testing requirements delays on the ring while of each end to perform link improving the efficiency quality testing, and to only slightly.[9] finalize the connection for Therefore, after extensive normal operation. Before performance modeling, the link establishment Digital decided to use enters its final phase, this default value for the the PHY types of the two requested token rotation ends of the connection time. The resulting network are compared, and the operates with low delay connection is allowed and high bandwidth as if the end types conform the default and does not to specific connection need complicated network rules. These rules are management procedures carefully established to achieve this level of to ensure that rings are operation. configured correctly and to Several unique but harmful prevent miscabling, which conditions in FDDI stations could cause partitioned must be addressed. Among and unnecessarily wrapped them are the prevention of rings. An established link and protection against is continually monitored duplicate addresses, for errors indicated by no-owner frames, and the reception of improperly the stripping of frames encoded data and is shut sent by bridges or down if those errors end stations that have exceed a predetermined multiple addresses.[10] threshold.[8] Because several of the After the PHY connection algorithms fundamental is fully established, the to the operation of the station's MAC is inserted FDDI technology use the into the ring, and the stations' addresses, claim process begins. the presence of two Digital's stations all use stations with the same a default requested token address causes numerous rotation time of eight malfunctions, ranging from milliseconds in their claim beacon-claim oscillations process to ensure a token to blocked communications rotation time, and hence between stations. Frames traffic latency, similar to not properly stripped from that experienced in other the ring can circulate LANs like Ethernet. FDDI forever, flooding the provides high aggregate stations that copy these bandwidth and, thus, frames. To protect against delivers the low delay strip errors, Digital's 8 Digital Technical Journal Vol. 3 No. 2 Spring 1991 Fiber Distributed Data Interface Overview chip set has several built- Before focusing on chip in mechanisms. Digital also development, Digital greatly improved the data carried out a large integrity of the ring. This simulation effort to ensure improvement is particularly the ANSI standards were important to token ring correct and complete. architecture where messages Once the standards were traverse virtually all of verified, modeling was the links in the network performed to produce before arriving at their chips that met these destinations. Robustness standards. Allowances and in the face of link bit trade-offs were made for errors becomes extremely unfinished sections and important. Digital designed future standards migration. several improvements to Real products were then the basic FDDI algorithms, planned around the chips and the ANSI committee as the use of FDDI was adopted them to improve the threaded into Digital's undetected error rates on a network architectures network.[11] and existing products. An implementation strategy Role of FDDI in Digital's for each product was then Networks formulated to balance the risks, the resources, and The FDDI technology is the timeliness of customer more than just another needs. data link interface It was important for that allows the use of Digital to understand transmit and receive fibers migration of current between devices. Digital's products and to think decision to embark on forward to the needs beyond the FDDI development the initial program goals. effort was a major program LAN management and host undertaking involving the connections take the FDDI development of VLSI chips technology beyond a simple and, subsequently, FDDI high-speed backbone for software and hardware the extension of bandwidth- products. Although the limited existing LANs. development of chips may Because Digital considered seem to be at the heart of the FDDI technology beyond the FDDI program, chips are its use as just another new certainly not the products data link, this technology that help customers solve is the natural choice as problems. Chips comprise the next step in network only a small portion of evolution. each large printed circuit board, but this portion is At the onset of Digital's an important one. FDDI program, and at several points during its development, a number of key program-wide decisions and policies were adopted. The commitment to Digital Technical Journal Vol. 3 No. 2 Spring 1991 9 Fiber Distributed Data Interface Overview simulation, both in breadth shipment, EAROM was (e.g., chip design, board included in all products design, and software) and to allow firmware to be extent (e.g., behavioral updated remotely over the models, gate-level models, network. As a result, EAROM integration of operational reduced the cost of product software with simulation enhancement by eliminating models) was essential to the need to change ROM the success of the FDDI in the field or swap out technology development. All boards and rework them at the chips developed were the factory. fully functional in their first pass in silicon, Deployment of FDDI and the integration of the controlling and test As a baseline effort in software with the chips was deploying FDDI in products, fast and smooth. Digital developed the FDDI Given the extensive design corner shown in investment in ensuring Figure 6, consisting of correctness of the FDDI chips and FDDI control technology, all FDDI firmware for use in all products were to use the applications of FDDI. same FDDI building blocks, In addition to using the including the chips and design corner in all of controlling software.[12] Digital's products, two Such sharing and reuse of of the chips critical to investment eliminated the FDDI interoperability, the duplication of effort, MAC and the ELM, together guaranteed the consistent with the SMT firmware, operation of all products, which controls the chips and shortened the time and the behavior of the to market. To ensure an station, were licensed even broader use of this to Advanced Micro Devices technology, major portions and to Motorola, Inc. for were made available on the manufacture and sale on the open market. open market. Widespread availability of this Another key program technology will foster decision was the adoption competition to drive down of electrically alterable cost and increase levels read-only memory (EAROM) of interoperability and in place of traditional consistency among FDDI read-only memory (ROM) to implementations. store the firmware in each product. SMT, which is implemented almost entirely in firmware, was a rapidly evolving specification while the products were under development. To accommodate these ongoing changes and the threat of change after product 10 Digital Technical Journal Vol. 3 No. 2 Spring 1991 Fiber Distributed Data Interface Overview FDDI Concentrator FDDI-To-Ethernet Bridge To make the FDDI technology The Ethernet/802.3 bridge widely acceptable as to the FDDI network is the next generation LAN, an important device several of the logistical in the first product obstacles inherent in set offered for FDDI a ring topology had to connectivity. The DECbridge be overcome. Connecting 500 product provides new stations to a LAN the smooth migration of can be frequent events. all Ethernet and IEEE These events must not 802.3 devices currently be restricted or cause a in the marketplace to disruption if the LAN is the increased backbone to be truly usable. The bandwidth of the FDDI nature of a ring topology technology. Most local causes the network to networks are LAN-based break whenever a station is with the LANs being removed or while a station interconnected to form is being inserted. The extended LANs by the use dual ring accommodates one of bridges. The bridges are such event by wrapping at transparent to all users' the adjacent stations, but protocols on the LANs. This two or more such events capability for transparent are disastrous because interconnection is one of they partition the network the keys to the instant into isolated parts. A use of and easy migration concentrator can solve to the added bandwidth and this problem by connecting other benefits of FDDI. to a station, testing the However, providing the link between the station transparent interconnect is and the concentrator, difficult. Frame formats, and then splicing the new bit ordering, padding station into the ring with fields, and even the length minimal disruption. At the of the packets differ same time, a concentrator between the Ethernet and can automatically drop IEEE 802.3 LANs. These a station out of the differences force the ring should the station frames to be translated malfunction or lose power. as they pass through the In addition to the obstacle DECbridge 500 device. This of connecting new stations, translation, coupled with it is inconvenient to a stiff requirement for map a ring topology into data integrity for the star wiring, which is MAC bridging standard well known and widely IEEE 802.1d and the used for its convenience performance needs of and manageability. dealing efficiently with Concentrators can turn packet rates approaching a logical ring into a 500,000 per second, hierarchical star-wired created a formidable design network, thus solving this task.[13] mapping problem. Digital Technical Journal Vol. 3 No. 2 Spring 1991 11 Fiber Distributed Data Interface Overview Digital's DECconcentrator applications and on the 500 product was built to network configuration. support simple installation One application is the and network configuration, use of the FDDI technology existing building cabling for a work group of high- plants, and the cost performance workstations savings and simplicity of and servers. Some high- single attachment stations. performance I/O-intensive This FDDI concentrator applications need more is low cost and flexible, bandwidth than a large, with a dual attachment shared Ethernet can and up to eight master handle. As a result, ports for connecting single a work group may form attachment stations or around an FDDI network. additional concentrators One application example is into the ring. With the called visualization, where concentrator, any of the a large volume of real- described FDDI topologies time graphics or imaging can be created to best information is transmitted suit a customer's cabling, over the network from a network management, and compute server. Various availability needs.[14] applications of digital image transmission have FDDI for Workstations few real-time latency requirements, but Digital designed a simple still require a large high-performance FDDI bandwidth.[17] adapter for TURBOchannel Another application of workstations and servers the FDDI technology in in conjunction with the workstation environments is design of the DECstation to achieve higher bandwidth 5000 reduced instruction for server/server and set computer (RISC)-based backbone interconnection. A system.[15] The aim of shared Ethernet for client this product was to produce /server as well as server an FDDI interface that /server communications can was both inexpensive be overloaded, particularly and high performance. if the population of With this adapter and servers or workstations ULTRIX operating system is large. FDDI can be used support, Digital attempted as a backbone to provide to address many of the higher bandwidth between shortcomings that have the servers. Alternatively, plagued most adapters the network may be a and operating systems hybrid, where the FDDI attempting to achieve high network is used for high performance.[16] client/server bandwidth by The applications of some systems and for high the FDDI technology server/server bandwidth for workstations and by others. Hybrid networks servers depend on the often have multiple FDDI particular choice of user networks and multiple 12 Digital Technical Journal Vol. 3 No. 2 Spring 1991 Fiber Distributed Data Interface Overview Ethernet networks. This kilometers such as a campus configuration is typical of backbone. Other medium evolving extended LANs, as types are currently under new technology such as FDDI investigation to provide is added into the existing optimization in areas network infrastructure. such as cost and ease of installation. FDDI in the Future As a new technology, Acknowledgments FDDI has a long future We would like to thank with opportunities for the efforts of the greater levels of circuit Telecommunications and integration, lower cost Networking LAN Architecture designs, and alternative and Development teams. media types to match the Without their hard work wide range of possible none of the many technical applications. Significant advances made in bringing progress has already been the FDDI technology made in this direction to market would have with the introduction been possible. We would of the DECcontroller 500 also like to thank Bob adapter, the most compact, Krueger for his role in least expensive, and one successfully leading the of the highest performing entire program. Finally, workstation adapters in we would like to thank Mark the industry today. In Kempf, Paul Koning, and addition, the ground- Tony Lauck for their timely breaking technical work review of this paper. and standards committee activity regarding Reference various types of copper media offers tremendous 1. N. Kronenberg, H. opportunities for cost Levy, W. Strecker, reduction which will enable and R. Merewood, "The broader utilization of the VAXcluster Concept: performance offered by the An Overview of a FDDI technology.[18] Distributed System," In addition to the Digital Technical original physical medium Journal vol. 1, no. 5 dependent standard already (September 1987): 7-21. specified by ANSI for FDDI, 2. R. Grow, "A Timed Token an additional standard Protocol for Local Area has been developed for Networks," Proceedings operation on single-mode of IEEE Electro/82 fiber over distances Conference, Boston, up to 30 kilometers. MA (May 25-27, 1982). This standard will be particularly useful in deploying an FDDI network that must span one or more distances greater than 2 Digital Technical Journal Vol. 3 No. 2 Spring 1991 13 Fiber Distributed Data Interface Overview 3. Token Ring Physical 10.H. Yang and K. Layer, Medium Dependent Ramakrishnan, "Frame (PMD), (International Content Independent Standards Organization, Stripping for Token reference no. ISO 9314- Rings," Proceedings 3, 1990). of the ACM SIGCOM '90 4. Token Ring Physical Symposium (1990): 276- Layer Protocol, 286. (International Standards 11.R. Jain, "Error Organization, reference Characteristics of no. ISO 9314-1, 1989). Fiber Distributed Data 5. Token Ring Media Interface (FDDI)," Access Control (MAC), IEEE Transactions on (International Standards Communications, vol. 38, Organization, reference no. 8 (August 1990). no. ISO 9314-2, 1989). 12.P. Ciarfella, D. Benson, 6. H. Yang, B. Spinney, and D. Sawyer, "An and S. Towning, "FDDI Overview of the Common Data Link Development," Node Software," Digital Digital Technical Technical Journal, vol. Journal, vol. 3, no. 3, no. 2 (Spring 1991, 2 (Spring 1991, this this issue): 42-52. issue): 31-41. 13.R. Kochem, J. 7. FDDI Station Management Hiscock, and B. Mayo, (SMT) Preliminary "Development of the Draft Proposed American DECbridge 500 Product," National Standard, ANSI Digital Technical X3T9/90-X3T9.5/84-49, Journal, vol. 3, no. REV 6.2 (May 1990). 2 (Spring 1991, this issue): 53-63. 8. J. Hutchison, C. 14.W. Tiffany, P. Koning, Baldwin, and B. and J. Kuenzel, "The Thompson, "Development DECconcentrator 500 of the FDDI Physical Product," Digital Layer," Digital Technical Journal, vol. Technical Journal, vol. 3, no. 2 (Spring 1991, 3, no. 2 (Spring 1991, this issue): 64-75. this issue): 19-30. 9. R. Jain, "Performance 15.U. Sinkewicz, C. Chang, Analysis of FDDI L. Palmer, C. Smelser, Token Ring Networks: and F. Templin, "ULTRIX Effect of Parameters Fiber Distributed and Guidelines for Data Interface Setting TTRT," DEC-TR Networking Subsystem 655 (Maynard: Digital Implementation," Digital Equipment Corporation, Technical Journal, vol. September 1989). 3, no. 2 (Spring 1991, (internal) this issue): 85-93. 14 Digital Technical Journal Vol. 3 No. 2 Spring 1991 Fiber Distributed Data Interface Overview 16.W. Hawe and K. Image Applications," Ramakrishnan, "The Proceedings of the Workstation on the Fifth Annual Phoenix Network: Performance International Conference Considerations for on Computers and the Communications Communications, Interface," Proceedings Scottsdale, AZ (March of the Second Workshop 26-28, 1986). on Workstation Operating 18.S. Ginzburg, W. Mallard, Systems, Asilomar and D. Newman, "FDDI Conference Center, CA over Unshielded (September 1989). Twisted Pairs," IEEE 17.K. Ramakrishnan and Proceedings, Eighteenth W. Hawe, "Performance Conference on Local of an Extended Local Computer Networks Area Network for (October 1990). Digital Technical Journal Vol. 3 No. 2 Spring 1991 15 ============================================================================= Copyright 1991 Digital Equipment Corporation. Forwarding and copying of this article is permitted for personal and educational purposes without fee provided that Digital Equipment Corporation's copyright is retained with the article and that the content is not modified. This article is not to be distributed for commercial advantage. Abstracting with credit of Digital Equipment Corporation's authorship is permitted. All rights reserved. =============================================================================