Mark F. Kempf,
Senior Consulting Engineer,
FDDI Program Technical Director
In the early 1980s, Digital introduceds local area network (LAN) products based on industry-standard, 10-megabit/second Ethernet technology. These products allowed microcomputers, terminal servers, and other network devicesto be connected with ease and offer unprecedented bandwisth. As local networks grew in size and began to strain the capabilities of a single LAN in the mid 1980s, Digitqal introduced products based on the extended LAN cocept. By connecting multiple LANs with filtering and forwarding bridges, a much larger number of stations could be interconnected with greated aggregate bandwidth.
A few years aftet the introduction of extended LANs. it became apparent that once again the capabilities of existing LANs wer being straines, but this time in two dimensions. Not only was there need for greater bandwidth in the "backbone" of the network, but with more than 10-megabits/second to a single station. The current and projected performance of workstations indicated that any new generation of LAN sutiable for future Digital products would have to offer an order of magnitude increase in bandwidth delivered to a single point. It was also clear that it would have to adhere to a widely accepted industry standard, since users had come to expect the abiliity to interconnect equipment from many vendors. After considerable analysis, Digital selected the emerging ANSI FDDI (fiber distributed data interface) 100-megabit/second token ring standard as the basis for our next generation of LAN products. The standard met our important requirements and showed promise of becoming widely accepted (today, of course, it is). However, because the standard was incomplete, our plans had to accomodate the implementation of specifications that would been implemented, we had to expect to discover errors in the existing document and to work with the standards organization to correct them. The first group of articles in this issue of the Digital Tehchnical Journal describes how we implemented the standard and some of the techniques we applied to deal with change as the standard matured.
Of course, in Digital's business, the implementation of a LAN technology is only a means to an end. The LAN technology must be incorporated into products that provide useful services to our customers. One important product in Digital's initial FDDI offering, the DEC FDDIcontroller 700 workstation adapter, meets the need for more bandwidth delivered to a single station. But LAN users have come to expect more: manageability, interoperability, and a generally radial wiring scheme; and they have large existing LAN infrastructures. Therefore, it was essential to introduce FDDI with a set of products that addressed all these needs as well. Digital's FDDI wiring concentrator, the DECconcentrator 500, permits highly reliable, manageable, radial wiring schemes. DECelms, extended FDDI and preexisting extended LAN products. It is important to remember that Digital does not view FDDI as a replacement for Ethernet, but rather as its complement. The large number of existing Ethernet LANs can be connected to FDDI using another element in Digital's initial FDDI product offering, the DECbridge 500 FDDI/Ethernet bridge. This product allows multiple Ethernets to be interconnected using FDDI as a backbone. Ethernet stations can communicate directly with stations on FDDI as well as stations on other Ethernets. The secdond group of articles in this issue describes how we developed these FDDI products.
During the entire development process, we relied on several design strategies. One emphasized he importance of designing the entire system that comprises the product set, not just individual products. Another was the decompsotion of problems. Early in the design process, we identified functions that appeared to be independent of others. We took into account how each component that implemented these functions would be used in the various products and how the products would interact with each other. In our development plans, we assumed that once a to work as expected when other components were added around it. Naturally, we could not assume that our initial assessment of independence was infallible, and we remained alert for unexpected interactions. Another related design strategy was partitioning to reduce risk. As mentioned before, the FDDI standard was evolving and subject to change. Therefore, we physically partitioned our design to closely mirror the separate sections of the standards document. This approach helped limit to a single component the impact of standards changes. The final important design strategy can be summarized as "analyze everything you can, simulatge what you can't analyze, and prototype what you can't simulate." The work, expense, and recovery time from errors increases with eash step through this progression, and the advantages are obvious.
Of course it is also obvious that this summary is simplistic since it is impossible to analyze, simulate, and prototype exhaustively. It is necessary to step into the gray area of risk assessment an engineering judgment to make satisfactory progress. FDDI LAN technology phases. In the first phase, we concentrate on analysis and simulation on the underlying FDDI algorithm specified by the standard. We wanted to ensure they actually described mechanisms that would produce a reliable, high-performance LAN. During this phase we also implemented the standard in silicon and software, and combined these components to form complete FDDI test stations.
Sinc standards documents, like all written documents, are subjedt to interpretation, two separate teams were give tne task of implementation and verification. We used the test stations, first in a simulation environment, to execute test scripts developed directly from the standards compliance. The scripts were also used to verify details specific to our implementations and had the additional benefit of making regression testing after imcremental changes relatively easy.
In the second phase we conncentrated kon proving thaqt significant numbers of FDDI stations, which had previously been shown to work individually, could be interconnected to form large reliable rings. This was the most important applicaiton of the heavily interactions between large numbers of asynchronous stations overwhelm both analytical techniques and available computing power.
In the third phase, the focus was on producing products. From a logic design and a software standpoint, Digital's FDDI products are largely derived from the test systems used in the previous phases. Of course, laboratory test capabilities were removed and major changes in power and packaging made; but this qapproach significantly reduced both the opportunity for introducing new errors amd the time to market. We were also able to use the capabilities of the laboratory versions to help verify the correct function of the products.
Numerous other activities that contributed to the effectiveness and timely delivery of the products were carried out simultaneously with engineering design. For instance, Digital maintained a significant presence at the FDDI standards committee to apprise the committee of various technical problems we found in the standard, to offer solutions, and to ensure our implementation reflected the intent of the standard. In addition, the close working relationships fostered between various organizations , expecially between development and manufacturing, resulted in products with a good balance between time and to market, function, performance, amd manufacturing cost.
The articles in this issue of the Digital Technical Journal go into much greater detail on the subjects I have touched on. I hope you find them interesting and informative.