taylor@ecsvax.UUCP (Steven Taylor) (07/30/85)
The following are a report description and a sample summary report for a report series which my company, Distributed Networking, is considering producing. These reports will provide the data communications industry with the first in depth, technically competent analysis of the wide area networking market. I would very much appreciate the net.dcom readers reviewing this information. In particular, I am looking for feedback about the feasibility of such a study, especially concerning the marketability once it is completed. Consequently, your comments from a technical viewpoint and from a potential purchaser's viewpoint would be most valuable to me. I also have enclosed a sample "management report" about one of the packet switching systems. This summary report is an actual preliminary evaluation of a real network, but the names of the network and the products have been changed to protect the interests of all parties. Thank you in advance for your help. Steven Taylor Distributed Networking Associates 119 Doncaster Lane Charlottesville, VA 22901 (804) 979-0656 ---------------------------------------------------------------------- TECHLINK Reports on Data Communications Networks Distributed Networking Associates' TECHLINK Reports on Data Communications Networks provide an analysis of networking systems in previously unavailable technical detail. The initial set of reports covers two of the most popular types of networks - high speed "T-1" networks and private packet switching netowrks. The TECHLINK Report on T-1 Networks examines the technology recently made available to end users to utilize high speed digital transmission facilities. Systems included here are primarily high speed multiplexing networks designed to carry voice and data over a wide variety of facilities such as telephone company T-1 facilities, private microwave, and private local area facilities. The TECHLINK Report on Private Packet Switching Networks examines systems commonly referred to as X.25 packet switches, distributed data switches, and switching multiplexers. The reports have a primary goal of assisting the user to determine which networking system, if any, best fits his organization's unique needs. As such, TECHLINK significantly reduces the time needed for evaluation of networking systems by providing detailed information about the major systems currently in the marketplace. Essentially, TECHLINK provides the user with the equivalent of a detailed consulting report on the state of the art in this technology at a fraction of the price that a "dedicated" consulting report would cost. The reports are highly tutorial in nature, helping the user make sense of all the trade jargon which often makes the data communications field somewhat difficult to understand. Because the TECHLINK reports are prepared by a vendor-independent consulting firm, they provide objective third-party technical analysis. Report Description Each TECHLINK report consists of four basic sections. The opening section is a detailed technical specification of the "ideal" system, coupled with an explanation of why each part of the specification is important. The explanation not only assists the report user in understanding the vendor responses, but also helps cut through the technical nomenclature by clearly defining all terminology. Consequently, this section serves the dual purposes of setting a framework for the remainder of the report and providing users with a set of specifications which may be customized for use in their own procurement procedures. The second major section contains summary "management reports" detailing the strongest and weakest features of each system, together with other commentary about each system's unique capabilities. This section is a quick reference guide in which it is assumed that most of the standard features found in such a network are provided and only the exceptional capabilities are noted. The results of the technical specification for each vendor constitute the third major section. This section is more than a typical vendor response to a questionnaire. Each vendor will be invited to have extensive conferences with the Distributed Networking Associates staff in the completion of the questionnaire to ensure accurate responses. Additionally, the completed text of the questionnaire will be presented to the vendor prior to publication for verification of all information. Finally, since there may be outstanding issues on which the Distributed Networking Associates staff and the vendor have professional differences of opinion, a fourth section is reserved for the unedited remarks of the vendor. The vendor may choose to include pricing information in this section, as pricing will not be discussed in the body of the report due to the tendancy for relatively frequent price fluctuations in this market. The vendor may also wish to include comments about field support for the product in this section. Scope of Report TECHLINK is planned to include reports on most major private network vendors. Specifically, the TECHLINK Report on T-1 Networks is planned to include T-1 multiplexing and networking systems offered by at least (alphabetically) Avanti, Bayly, Case Rixon, Codex, Datatel, DCA, Gandalf, General Datacom, Infotron, M/A-Com Linkabit, Paradyne, Racal-Milgo, Spectrum, Tellabs, and Timeplex. The TECHLINK Report on Private Packet Switching Networks is planned to include systems offered by at least (alphabetically) BBN, Case Rixon, Codex, ComDesign, DCA, Develcon (Develnet), Doelz Networks, Dynapac, Equinox Systems, Gandalf, Infotron, Micom, Network Products, Northern Telecom, Paradyne, Penril, Siemens-Databit, Telematics, Tellabs, and Timeplex. Additional Services Additional support services regarding this report are available to purchasers of the reports. In particular, there are plans for quarterly updates to the reports. These updates are an essential part of the product due to the rapid changes in the data communications industry. The updates will add new information about systems in the report and add entire new systems as these systems are announced. Communications systems consulting regarding customized areas of the reports will be made available. Seminars, both public and in-house, also will be available to help educate staff in the use of the reports. Price and Availability The initial release of TECHLINK is planned for publication during the first quarter of 1986. Quarterly updates will follow. A special prepublication price $895.00 per report or $1495.00 for both reports is offered for orders received prior to October 1, 1985. The first year of update service may be ordered in advance for $200.00 per report per quarter. Customers who order both the basic report package and first year's update service will receive a 50% discount on conference fees for seminars detailing the results of each report. Additional in-house seminars and consulting will be priced separately. Distributed Networking Associates Staff Distributed Networking Associates is involving a consortium of consultants, educators, and end users in the preparation and editing of this report. These individuals constitute an "Industry Advisory Board" which is involved in all aspects of the report preparation. A "Manufacturers Advisory Board" is also providing vendors with a mechanism for recommending that certain areas be given special attention in the specification and evaluation process. This unique blend of users and vendors should enable the professional quality of the report to yield extremely high quality information, both in asking the right questions and in obtaining accurate answers to the questions. ----------------------------------------------------------------------- Management Overview - "Generic" Networks This report provides a brief technical description of the networking product offered by "Generic" Networks. In general, the product is very similar to most packet switching and switching multiplexer products which are currently on the market (or at least announced). However, there are four areas in which this product is sufficiently unique that special attention is required. These areas are the specialized link protocol, the ability to handle multidropped lines, the ability to support any synchronous protocol without a specialized handler, and network management capabilities. There are two basic hardware versions. The "Mini-Net" is the smaller product, consisting of up to two composite high speed data links per node and up to 32 ports. Data links are rated at up to 72 kbps. The "Maxi-Net" is the larger system, and calls for support of up to 67 data links to "Mini-Net" systems and up to 8 data links to other "Maxi-Net" systems. Composite data link speed is also 72 kbps for the "Maxi-Net," although 1.544 mbps is planned. The "Maxi-Net" also features a quite elaborate redundancy structure, including redundant data buses. Specialized Link Protocol The "Generic" data link protocol is indeed a fairly radical departure from the way that packet switching traditionally has been approached. At the port level, the data from each individual session is packetized in "small, fixed-length" packets. These packets are similar to X.25 Level 3 packets. Error checking, if desired, occurs end to end only. (Window size was not immediately available.) The data links in a statistical multiplexer or packet switch are typically a variation of X.25 Level 2. As such, they are optimized for point to point operation. They also perform error checking on the data link level. The "Generic" products don't really use a Level 2 protocol at all. Instead, they use a combination of stat muxing and local area network architecture. When data from a particular port is ready to be placed on the network, the packetized data, with its network address, is placed on the outgoing circuit (transmit data leads.) At this point, a "ring architecture" takes over. Imagine a "daisy chain" of about five "Mini-Net" nodes. ("Generic" calls this a "network link.") The middle node receives some data on a port and places this on the "transmit data" pair. The next node receives the data on its "data link 1." If the data belongs to that node, the data is shipped off to the appropriate port. If the data does not belong to a port on that node, the entire packet is immediately forwarded to the transmit pair of "data link 2." This process continues to the end of the chain, where all data not destined for the end node is turned around and retransmitted. "Generic" believes that this architecture is extremely efficient because the data is not reframed at each intermediate node for the Level 2 protocol. This efficiency is realized primarily because less processor power is used in each node for reframing the data. There are some technical concerns about the additional loading on the data links caused by the ring architecture. This, combined with the short packets, could result quite significant overhead. The architecture allows for rerouting of calls in the event of a failure. This rerouting uses a primary circuit and a secondary circuit for each "network link." A standby circuit links the endpoint of the "normal" network link with the saxi-Net" system. The major change is that each endpoint on an "Maxi-Net" terminates a "network link," and the "Mini-Net" may then pass data among multiple "network link" rings. "Maxi-Net" to "Maxi-Net" communication consists of a simple two-point ring. At this point there may be only one "Maxi-Net" per network link. Multipoint Support The "Generic" protocol described above enables the support of multidropped terminals. At this time, however, all multidropped terminals on a single "virtual multipoint circuit" must be on the same network link. This capability is quite unique within the packet switching marketplace. In environments where multidropping is very firmly entrenched, such as the banking systems using ATM's, this is indeed an attractive feature. In all fairness, however, it should be mentioned that there are questions about the advisability of multidropping in many situations from both technical and financial viewpoints. "Any" Protocol Support The "Generic" approach of using short, fixed length packets to minimize throughput delay is key to their claim of being able to support any protocol, including bit synchronous protocols such as SNA/SDLC. The delays inherent in the multiplexing are compensated for by the use of a fixed duration delay timer. Consequently, there are no protocol-specific data handlers. The throughput for synchronous channels is further enhanced by the utilization of "selective ARQ." The selective ARQ enables or disables the ARQ function on a channel by channel basis. Since most synchronous protocols have inherent error detection and retransmission capabilities, the error checking by the network is not needed. Thus, in bypassing this step, the "Generic" product decreases the delay which would be encountered while retransmitting blocks of data containing transmission errors. At this point, host/terminal emulation, or "spoofing," is not a part of the synchronous capabilities. All data is passed through the network. "Generic" reportedly has an "auto-emulator" under development, however. The "auto-emulator" would decipher the normal polling pattern and then emulate the pattern. Like a currently available emulator, the "auto-emulator" would only pass data and would not pass polls. However, because of the "self-learning" process, the emulator would not have to be specific for each protocol. This approach is indeed interesting. The basic idea seems to be that if the data is delivered fast enough, no protocol-specific handlers are needed. However, with high levels of congestion of the data links due to peak traffic periods, deteriorating communications facilities, etc., problems such as device timeouts may be experienced much more frequently than when emulation is used. The real question, consequently, is whether the price paid for enough bandwidth to ensure "fast enough" delivery is economically advantageous over the price paid for protocol-specific emulation. "Generic" partially addresses the problem of delays when there is heavy bandwidth contention by allowing the network manager to assign priorities to each session. Consequently, it is possible to help preserve synchronous device response assigning a higher priority to synchronous devices than is assigned to asynchronous devices. This approach is fine so long as better response is indeed desired for synchronous devices than asynchronous devices in the network under consideration. In some network operations, however, it is desirable to have better response on interactive aysnchronous terminals than on synchronous batch terminals. One final philosophical note on this issue. The "Generic" approach relies on using the error checking mechanisms of the synchronous protocol for error detection and retransmission. This indeed represents a significant savings over "dual error correction" seen by using methods which pass all of the synchronous data over error checked data links. However, the savings in terms of decreased delay, increased compaction ratios, etc, have not been demonstrated versus a system which uses protocol-specific emulation. Here, one would expect issues such as block length, number of intermediate nodes, etc., also to have a severe impact on the determination of which method provided the better performance. Network Management Network management is one of the strongest features of the "Generic" network. "Generic" estimates that up to half of the processor power utilized in the system is devoted to network management tasks. While insufficient data is available to evaluate fully the merits of the management system in this preliminary report, the fact that network management is designed as an integral part of the system instead of an afterthought provides significant advantages. Summary and Primary Applications The "Generic" networking products offer some significant new ideas in handling synchronous protocols. While the intent of the network is to remain transparent to all protocols, it remains to be seen whether this can be accomplished while still providing acceptable performance under stressed data link conditions. The intended transparency also significantly limits the ability to offer protocol-specific services, such as switching (selection) functions for synchronous devices. Major advantages include the ability to support large networks, protocol-independent synchronous transmission, multidrop (multipoint) support, failure compensation, and network management. Possible limitations include unproven performance on stressed data links due to overhead considerations and delays in synchronous data and questionable efficiency of bandwidth utilization on "network links" due to the ring architecture. These advantages and limitations will probably make this network most attractive to users with relatively large networks utilizing multiple data links at speeds of 56 kbps or greater. Greatest advantages will also be found for networks with a large percentage of interactive synchronous devices, such as ATM's and 3270-type terminals, which require good response time but have relatively light absolute network loading.
sjl@amdahl.UUCP (Steve Langdon) (08/03/85)
> > The following are a report description and a sample summary > report for a report series which my company, Distributed Networking, > is considering producing. These reports will provide the data > communications industry with the first in depth, technically > competent analysis of the wide area networking market. > ... > I also have enclosed a sample "management report" about one of > the packet switching systems. This summary report is an actual > preliminary evaluation of a real network, but the names of the > network and the products have been changed to protect the > interests of all parties. ^^^^^^^^^^^ > > Thank you in advance for your help. > > > Steven Taylor > Distributed Networking Associates > 119 Doncaster Lane > Charlottesville, VA 22901 > (804) 979-0656 All parties except the net.dcom readers who are curious about which company makes the product described in the report. For those who do not recognize the product, the company selling the network described is Doelz Networks Inc. -- Stephen J. Langdon ...!{ihnp4,hplabs,sun,nsc}!amdahl!sjl [ The article above is not an official statement from any organization in the known universe. ]