Due to the increasing infromation capacity of optical fibernetworks the need for immediate restoration has become a necessity. The fiber cut or `back-hoe fade' is now a significant factor to network design and implementation. In this paper details of network restoration techniques relying on diverse fiber routing and intelligent optical switches are given.

In this paper, we present the architecture and components of two experimental multimedia communication systems built by integrating available hardware devices and developing software modules. These platforms are used to investigate various issues related to the development of distributed multimedia applications. The problems encountered during the implementation of these systems, as well as the lessons learned, are discussed. Some experimental performance measurements and major identified bottlenecks are given for each architecture.

In the following paper we discuss the architecture, performance, and applications of an ATM based SMDS switch. The core is an ATM switch with 8 X 8 switching matrix with OC- 3c (155.520 Mbps) speed. The SNI interface for SMDS operates at DS3 (44.736 Mbps) rates. At the user side a SMDS router converts the output of a LAN to SMDS format and transmits the cells over DS3 to the switch. The first applications are linking computers connected to Ethernets and FDDI rings. Traffic is monitored on an OAM terminal and a cell count is provided for billing.

An attempt is made to explain the totality of the experimetal data the author is aware of on optical properties of films of different thicknesses from supr thin to bulk single crystals obtained by various optical spectroscopy methods within the framework of macroscopic linear crystal optics in the polariton language.

One of the challenges for future internetworks is that the design should deal with ever-increasing diversity of speeds, access protocols, and resource allocations. In this paper, various performance issues for interconnecting (a) local area networks with high-speed LANs and (b) high-speed LANs with wide-area networks are discussed. Throughput, delay, and buffering requirements at the local and extended-area access controllers are analyzed. For simulation model, two architectural candidates, (a) FDDI and (b) token ring backbones interconnected with Ethemets, are considered. The simulation results include throughput versus offered load, load versus delay, and mean queueing delay at the bridges.

For the next generation telecommunications network, the CCITT has been developing the BISDN standards that will provide a new network service infrastructure. For the next generation high-speed customer premises network the ANSI X3T9.5 working group has begun to define a new standard, the FDDI follow-on LAN (FFOL). The compatibility between these two standards should benefit users. In order to analyze the interworking scenarios between FFOL and BISDN, this paper defines three possible FFOL architecture models that are synthesized from existing FFOL proposals. This paper also identifies key functions for seamless integration between FFOL and BISDN.

It is a commonplace observation that computational power at the desktop is increasing at an exponential rate. This continues two decades after the first single chip VLSI microprocessor became commercially available and it is projected to continue for at least another decade. As a direct consequence, several observations can be made about the revolutionary impacts occurring in data networking: (1) Inexpensive computer power has made it economically feasible to distribute immense computational capacity to the desktop. (2) Distribution has created a demand for sophisticated networks to enable resource sharing among work groups. (3) Placing compute capacity at the point of consumption has removed the communication barrier from the `man/machine' interface. Virtually every user of computer systems is presented with increasingly rich visual paradigms. Current graphical user interfaces are designed to take advantage of bit mapped color displays that have spatial resolutions of 1024 pixels X 1280 pixels and 8 to 24 bits per pixel of color resolution. (4) Standards have been defined and systems are being built to extend the visual paradigm over the networks that interconnect information workers. (5) As a result of the exponential increase in computing capacity available for constant dollars, one would expect the demand networking capacity to increase accordingly. However, as a consequence of observation (4), the rate of increase is far greater. One of the narrow effects of the above has been to accelerate the demand for high performance networking solutions to support the burgeoning users of PCs and workstations. Fiber distributed data interface (FDDI) standard based bridges and routers have received rapid acceptance to provide backbone connections among Ethernet segments. It is not uncommon for an organization to have dozens of Ethernets within a single establishment. The cost of FDDI compatible interface boards for workstations and PCs is declining rapidly. This year the industry expects to see commercial availability of FDDI over twisted copper pair media that will further reduce the cost per connection. This should have the effect, over the next couple of years, of replicating the current multiple Ethernet interconnection problem but with multiple FDDI rings each operating at 100 Mbps. This paper is intended to explore one solution being developed to address what the author expects to be a large demand for the local interconnection of multiple FDDI rings. What is more, it is expected that the increasing trend toward `networked' applications will demand constant performance over networks that span the entire establishment if not an enterprise.

This paper presents a system architecture for a VMEbus FDDI adapter card containing a node core, FDDI block, frame buffer memory and system interface unit. Most of the functions of the PHY and MAC layers of FDDI are implemented with National's FDDI chip set and the SMT implementation is simplified with a low cost microcontroller. The factors that influence the system bus bandwidth utilization and FDDI bandwidth utilization are the data path and frame buffer memory architecture. The VRAM based frame buffer memory has two sections - - LLC frame memory and SMT frame memory. Each section with an independent serial access memory (SAM) port provides an independent access after the initial data transfer cycle on the main port and hence, the throughput is maximized on each port of the memory. The SAM port simplifies the system bus master DMA design and the VMEbus interface can be designed with low-cost off-the-shelf interface chips.

A design solution to implement an FDDI dual attachment station on a PC-AT interface card is provided. The implementation uses AMD's SUPERNET 2TM chip-set for the MAC and PHY controllers and results in a low cost and low power interface card for end station applications.

A station implementing the Accredited Standards Committee (ASC) X3T9.5 fiber distributed data interface (FDDI) physical layer communications protocol standard can transmit and receive data simultaneously at a data rate of 100 Mb/s through a pair of optical fiber cables or fiber channels. Each station, with multiple physical layers and their associated fiber channels connected together in parallel, can transmit and receive data at a much higher rate. National Semiconductor's physical layer (PHY) devices can be conveniently connected together to achieve a gigabit bandwidth point to point data link system, whereby the bandwidth is directly proportional to the number of PHY devices used.

A recent expansion of Argonne National Laboratory (ANL) forced an entire division, some 400 employees, to move to new off-site facilities approximately three miles from the main lab campus. This division is highly dependent upon computing and networking facilities at the lab that were not to be duplicated at the new facility. FDDI was chosen as the primary connection between the new facility and the lab. The distance is beyond the range of multimode fiber, so single mode fiber with appropriate adapters was utilized for the connection. Since the facility was new construction, fiber as well as copper was pulled to each office for future expansion.

This paper provides an overview of the U.S. standard (ANSI/EIA/TIA 568) addressing topology, distances, media (fiber and copper), and connectors. The impact of the wiring standard on new application standards, IEEE 10BASE-T, 4/16 Mb/s token ring, and FDDI is discussed showing how the wiring standard is evolving into a utility. The factors affecting the decision on pulling fiber to the desk are weighed against the ever-increasing data rates being delivered over copper at 100 meters (the maximum horizontal distance). The use of other media, including plastic fiber, are considered. The relationships between the emerging international wiring standard and ANSI/EIA/TIA 568 are examined showing the path toward a worldwide utility. The status of the United States, Canadian, and Australian federal government and international standards is reviewed.

The motivation behind this work is to propose a scheme that enables an easy (i.e., economical) upgrade of a communication system based on optical multimode fiber from a network operating at tens to hundreds of Mbps, such as FDDI, to a network capable of supporting Gbps transmission.

The design of the physical layer of a gigabit LAN is discussed. The interaction of synchronization techniques and architectures, signal encoding, and media characteristics is explored.

The `FDDI' designation is gaining an increasing place in the local area network (LAN) jargon along with the `Ethernet' and `Token Ring.' The four letters stand for fiber-optic distributed data interface, but as in the case of other LANs the meaning of the name looses its importance as the properties of the network become familiar to the community of the users, implementors, and designers. Further, more new properties are added to the original set, sometimes beyond the boundaries hinted by the name. This paper presents the actual stage of an attempt to change the `F' (fiber-optic) to `M' (metal) specifically to unshielded twisted pair (UTP). The sense in doing so is the fact that a huge installed basis of metallic wires for telephone and data transmission already exists, and reaches an immense number of desktops. This paper describes: (1) An hierarchical network architecture emphasizing the segment to be implemented over UTP. (2) A systemic approach to the definition of the parameters for the physical medium dependent (PMD) module that should interface the MDDI (FDDI over metallic media) to the UTP cable plant. (3) Measurement results available at the time of the presentation.

This paper describes the characteristics and performance of a high speed, burst mode compatible receiver for optical bus or packet communications. It employs a Si bipolar differential transimpedance amplifier, a threshold tracking control circuit, and a dc-coupled ECL compatible quantizer (decision circuit). To cope with intermittent data packets, the threshold control circuit can capture data amplitude and set the logic threshold in about 1 ns. Using an avalanche photodiode, the typical receiver sensitivity is -37.5 dBm (10-9 BER) at bit rates up to 900 Mb/s, with a dynamic range of 23 db for both pseudo-random and burst-mode signals. At 1 Gb/s, the sensitivity is -35 dBm. With a worst case reset time < 100 ns for the threshold control circuit, this receiver can be used for optical bus applications where data signals with varying optical power are employed.

A 266 Mb/s optical link card based on a 780 nm compact disc (CD) laser has been designed by IBM Rochester to provide a fast data link for interconnecting computer systems. This design not only met the performance and cost objectives of the data communications environment, it also overcame key challenges in the areas of laser safety, laser reliability, and modal noise. The 266 Mb/s data rate was selected to conform to the new American National Standard Institute (ANSI) Fiber Channel Standard (FCS).

Laser safety has become an important issue in the design of optical data links for computers because semiconductor laser diodes, rather than LEDs, are increasingly being selected as the light source. It is very important that optical data links in the computer environment meet the requirements of a Class 1 laser product (Class 1 is the safest, least restrictive laser safety category) due to the potential for user exposure to laser radiation. This paper presents an overview of an open fiber link detection and laser control system for point-to-point data links referred to as the open fiber control (OFC) system. The system functions by detecting whenever the optical link between two transceiver ports is opened (i.e., an open connector or cut fiber) and forcing the transmitters into a low duty cycle mode of operation that maintains a Class 1 output level from the open link.

A 2500 gate mixed signal gate array has been developed that integrates custom PLL-based clock recovery and clock synthesis functions with 2500 gates of configurable logic cells to provide a single chip solution for 200 - 1244 MHz fiber based digital interface applications. By customizing the digital logic cells, any of the popular telecom and datacom standards may be implemented.

Fiber channel standard (FCS) is a sanctioned development activity of the American National Standards Institute (ANSI) X3T9.3 working committee. FCS is essentially a high speed, low overhead, peer-to-peer serial transport mechanism primarily intended for transmission over fiber optic media. As such, FCS is rapidly gaining recognition as an important high performance channel for the interconnection of supercomputers, workstations, and I/O controllers. Although FCS was originally conceived as a data channel for the new generation of high performance computers ranging from workstations through supercomputers, it has incorporated facilities to support the network communication paradigm.

Fiber channel is a new interface being developed for data system communications. This paper discusses the aspects of the physical interface including the technology options and interoperability considerations. Emphasis is given to the flexibility inherent in the structure of the standard that allows it to adapt to many communications technologies and allows for future extensions.

The ever increasing speeds in data computation and communication have created a need for channel and interconnecting architectures with performance well beyond current capabilities. The services required for fast and efficient data transfer among high performance computing and peripheral devices are provided by the fiber channel compliant interconnection scheme, termed `fabric.' In addition to these services, this paper describes the services offered by the fabric to provide `concurrent data availability,' to the interconnected devices, required by some applications. The structure of the `distributed' fabric to provide services based on an inexpensive interconnection scheme is also reviewed.

This paper presents a brief explanation behind the drive for gigabit/sec communications for individual users. A description is given of a resulting prototype LAN that delivers this capability to each user. The prototype is being built for the Lawrence Livermore National Laboratory. It was to be delivered in late 1991 and should be available for users by the end of the first quarter of 1992.

The high performance parallel interface (HIPPI) is an emerging ANSI X3T9.3 Committee standard for high speed data communications over high performance computer networks. The maximum allowable cable length for transmitting the HIPPI-PH parallel data is 25 meters. The purpose of a fiber optic serial HIPPI extender is to provide a totally transparent, highly reliable connection up to 10 km in length. A means of serializing HIPPI data is described together with elements of the fiber optic design. The frame format plus the fiber optic component and link requirements are discussed. Performance of the BCP Model 1200 HIPPI extender is given.

An error correcting code has been developed for the serial transmission of digital data over a fiber optic link. It is a 96 bit block code, with single bit error correction and double bit error detection. Each block consists of eight 8b/10b coded symbols and sixteen error correction bits. The coded data has a maximum run length of 4, minimum transition density of 25%, and a digital sum variation of +/- 4 bits. A chip set implementing the code has been demonstrated in an experimental link running at 1 Gbit/s (1.5 Gbaud).

Providing a single user with a very high (on the order of gigabit per second) bandwidth is a challenge that the research community is trying to address these days. The current commercially available (or available in the near future) networks, such as FDDI, can deliver on the order of hundreds of Mbps to the end user. In some of these networks, the aggregate network bandwidth is much larger, but a single user can utilize only a small portion of the network capacity. We describe here a design of an `almost-all' optical local-area network that is capable of providing Gbps directly to the user. The network design is based on the 802.6 topology, the `field coding' technique that was reported by us earlier (in which the header and the data fields are encoded in different rates), and the principle of `almost all' optical switching. Several considerations guided us in our design, among them: cost, integration, and flexibility. By providing all of the above attributes, the optical distribution channel may be an example of future high-speed local area networks, in line with the FDDI-follow on (FDDI-FO) effort.

Several physical communication networks currently available can offer very high bandwidth. At the same time, many potential applications are in need for all that bandwidth. Unfortunately, the communication subsystems used today do not preserve the raw network throughput up to the application level. In this paper, we give an overview of the major current trends and approaches proposed to solve the so-called transport bottleneck problem. We discuss issues related to protocol design and implementation as well as the provision of new features at the transport service interface to accommodate emerging applications.

This paper introduces a collisionless wavelength division multiple access protocol for a passive star-coupled parallel computer. A performance analysis is developed based on a semi-Markov model. This model, verified by extensive simulation, is then used to study the behavior of the protocol with varying system characteristics.

This paper proposes a fast packet switching architecture based on time-division multiplexing a photonic space-division fabric. Asynchronous time-division multiplexing is applied to a wideband strictly nonblocking fabric to resolve destination conflicts and queue packets on outputs. High performance is achieved in terms of delay-throughput characteristics, packet loss probability, and switching overhead parameters. Performance and complexity are evaluated and results are analyzed.

Computer networks must provide high data transfer rates to maximize the effectiveness of the interconnected equipment, and especially to maximize the effectiveness of the users, e.g., with visualization. Network speeds are increasing, with the newest systems using 800 Mbit/s data rates. The most common computer networks today are bus and ring architectures. Supercomputer networks are starting to use circuit switching with crossbar switches. Wavelength division multiplexing and all-optical networking are research topics today, but hold promise for the future. The architectures, attributes, and problems of these different systems are discussed, with emphasis on their use in the supercomputer environment.

The prototype high performance switching system will connect several different types of machines with widely varying clock and internal bus speeds. At one end of the range we have a CRAY YMP and a CRAY XMP with a high performance parallel interface (HIPPI). At the other end there are many SUN 3/260 VME bus machines. In between there are various flavors of MAC IIs. This presents a substantial speed matching problem. The CRAY HIPPI channel speed is 100 MB/s. The SUN VME bus speed is approximately 1 MB/s with bursts to 13 MB/s, the MAC II NU-BUS obtains about the same speed. So what do you do when the network is 2 to 3 orders of magnitude faster than the internal bus? The short answer is use big packets and even bigger buffers. This paper expands on that theme and describes the approach LLNL and Ancor Communications took in the case of the SUN VMEbus machine.

In order to support existing and future applications, Argonne is installing a laboratory-wide fiber optic cable plant. It has been designed to support an integrated data/video service consisting of a laboratory-wide FDDI network and video teleconferencing system and has the capacity for future gigabit per second networks. This paper describes Argonne applications expected to push existing network capacity, our current network configuration, networking research activities, and future networking plans.

The superconducting super collider (SSC), a Department of Energy funded 8.25 billion dollar high energy physics laboratory currently under construction south of Dallas, Texas in Ellis County, will be the world's largest proton accelerator with a main ring 54 miles in circumference and a capability of 20 TeV. When completed in 1999 the laboratory will house several thousand support staff and researchers all interacting with one another and the collider by way of a networked distributed computing environment. It is estimated that during operation the collider's detectors alone will generate over 5 terabytes of data per day in addition to the many terabytes of data required for the ongoing operations of the laboratory. Currently, the network is a mix of copper and multimode fiber optic technology using async, ISDN, T1, Ethernet, and FDDI to provide support for administrative computing, engineering and design, simulation, and video conferencing over an extensive local and wide area network. As completion of the collider approaches and high energy physics experiments are begun the networks role will become ever more crucial with bandwidth demands at an all time high. To meet these demands the existing network will migrate to a mostly single-mode fiber optic system utilizing higher speed technologies such as T3, FDDI follow-on, SMDS, SONET, and serial HIPPI, to support the additional needs of data acquisition, control systems, and environmental control and safety systems. Throughout the design and implementation of the network several themes persist: `the network is the system,' bandwidth requirements are on the increase, solutions must be standards based, and fiber optics will prevail.

Local data delivery systems continually evolve to support the needs of their users. In many cases, LANs as they exist do not support many of the applications needed today or expected to be required tomorrow. This paper discusses how new and emerging LAN standards might evolve in a multiplexed environment to support integrated services and higher bandwidths.

High-resolution, digital image transmission challenges pale in comparison to the requirements of high performance interactive graphics between internetworked data sources and sinks. The situation is exacerbated by internetworking devices. Traditional network communication processing solutions are unable to address both high-throughput and low-latency product expectations. Protocol Engines incorporate new technologies targeted to address the emerging high-bandwidth and low latency expectations. These technologies are usable in data source and sink, as well as internetworking devices.

Multimedia communications systems are a combination of human interfaces and end users interacting with multimedia data bases and highly disparate but interconnected communications networks. This paper discusses several architectural alternatives and system requirements that will assist in the design and development of MMCS in actual environments. The approaches taken in this paper are based upon the development of such systems in both medical and printing and publishing environments. This paper develops several key concepts as how best to define and structure data in a multimedia environment, how best to integrate the communications elements, and how best to permit the maximum flexibility to the end user to utilize the system's capabilities in the context of a fully conversational environment.

The growing size and complexity of FDDI-based networks has made computer simulation the most effective tool for analyzing such networks. This paper demonstrates how the block oriented network simulator (BONeS) can be used to study the steady state performance of several FDDI LANs. The simulation studies are designed to show how various parameter settings and network configurations can affect delay, throughput, and utilization of such computer networks. The performance of the asynchronous priority scheme is studied in the presence of synchronous traffic. A simulation of a single-priority campus area network (CAN) using FDDI as a backbone is also presented.

This paper describes a high-speed communication link that serializes and transmits 40-bit-wide data over fiber at rates of up to 800 Megabits per second (Mbps). The medium can be either 50/125 (mu) or 62.5/125 (mu) fiber. The compact card containing hot rod and CD laser-based fiber optics can be used for serial links of up to 300 m at 800 Mbps, 1 km at 400 Mbps, and 2 km at 200 Mbps. The applications include high-speed networks and point-to-point communication.

A multigigabit/second local area network is described that uses a subcarrier multiplexed optical carrier to provide asynchronous, parallel information channels that can simultaneously transport variable-length datagrams, ATM-cells, and circuit-switched video traffic. A separate `token-passing' control channel provides efficient multiuser access to the information channels.