51-20-36 Bridging Strategies for LAN Internets Previous screen Nathan J. Muller Payoff As corporations continue to move away from centralized computing to distributed, peer-to- peer arrangements, the need to share files and access resources across heterogeneous networks becomes all the more necessary. The need to interconnect dissimilar host systems and LANs may arise from normal business operations or as the result of a corporate merger or acquisition. Whatever the justification, is becoming ever more important, and the interconnection device industry will grow for the rest of the decade. Introduction The devices that facilitate the interconnection of host systems and LANs fall into the categories of repeaters , bridges, routers, and gateways. Repeaters are the simplest devices and are used to extend the range of LANs and other network facilities by boosting signal strength and reshaping distorted signals. Gateways are the most complex devices; they provide interoperability between applications by performing processing-intensive protocol conversions. In the middle of this “complexity spectrum” are bridges and routers. At the risk of oversimplification, traditional bridges implement basic data-level links between LANs that use identical protocols; traditional routers can be programmed for multiple network protocols, thereby supporting diverse types of LANs and host systems over the same WAN facility. However, in many situations the use of routers is overkill and needlessly expensive; routers cost as much as $75,000 for a full-featured, multiport unit, compared with $6,000 to $30,000 for most bridges. The price difference is attributable to the number of protocols supported, the speed of the Central Processing Unit, port configurations, WAN interfaces, and network management features. Some vendors bundle selected functions of both devices into the same unit, permitting concurrent bridging and at a reasonable cost. Bridge and router applications are summarized in Exhibit 1. A Comparison of Bridge and Router Applications and Costs Bridge Applications Router Applications ------Previous screen Best for point-to-point Accommodate several data links and and simple mesh topologies. can exploit complex mesh topologies in cases of link failure and congestion. Easier to install and main- tain than routers. Support multiple network and router layer protocols at the same time. Operate independently of higher-level protocols. Offer advanced administration and control services based on network and Offer a flexible method for subnetwork addresses. filtering traffic according to source-destination add- resses, protocol type, and application. ------Cost: $6,000 to $30,000 Cost: As much as $75,000 ______

Although many companies are implementing LAN/WAN networks based on such vendor-neutral protocols as the transmission control protocol and Internet protocol (TCP/IP) and Open Systems Interconnection, these same companies have considerable investments in System Network Architecture equipment and applications. With many vendors providing connectivity to IBM Corp.'s Token Ring LANs and the SNA environment, it is worthwhile to review IBM's method of bridging and contrast it with the available alternatives. Connectivity Concerns IBM System Network Architecture is still a dominant architecture, but it is no longer able to meet the networking demands of the majority of users. Consequently, its host- controlled, hierarchical structure, with all information flowing through a central point, is rapidly being displaced by distributed computing and peer-to-peer networking over LANs. Although IBM very effectively addresses these needs with its Token Ring LANs, its preferred method for interconnecting them—source routing—for a variety of reasons is unsuitable for building large networks. Nor does source routing support other popular internetworking protocols (e.g., TCP/IP and Novell, Inc.'s IPX/SPX), making it unsuitable for multiprotocol networks. While users appreciate LANs for their efficiencies and economies, older SNA equipment still provides dependable service and may not yet be fully depreciated. Therefore, another problem faced by users is how best to eliminate parallel networks by integrating Synchronous Data Link Control and binary synchronous communications (BSC) serial SNA protocols with LAN traffic on a single network. Bridging Methods Several bridging methods are currently available: source route bridging, preferred by IBM; transparent bridging, a basic method of LAN interconnection supported by most bridge makers; and Source Routing Transparent, a relatively new standard that allows source routing and transparent bridging to be used together on the same network. Source Routing Previous screen Source route bridging is a method of internetworking Token Ring LANs that uses a process called route discovery to find the optimal path for communications between end stations. The route between end stations is discovered using “explorer packets” that are sent between the source and destination end stations. When the explorer packet reaches its destination, the end station responds by issuing a packet containing the Routing Information. If multiple routes are available, this packet is sent back to the source over all of the routes. The originating station selects as the best route the one with the fewest hops to the destination station. One problem with Source routing is that in mesh networks it creates a significant amount of overhead, which can bog down network performance. The amount of overhead increases as more stations and links are added to the network. Because the end-stations are involved in route selection , they may not have up-to- date knowledge of the best path, especially if the network is temporarily congested. In not being able to implement adaptive routing, Source routing bridges are not able to dynamically reroute traffic around failed links. To do this, a new route discovery sequence must be initiated. Source routing bridges also cannot balance the traffic load in response to congestion. Transparent Bridging Transparent bridging originated in the environment. It enables stations, regardless of location, to communicate as if they were on the same LAN. In a process called filtering, the bridge looks at the destination address to see if it is listed in the table of source addresses. If not, the packet is sent over the bridge to the next LAN. If a match is found, the bridge simply ignores the packet. In a process called learning, the bridge examines all the packets originating on the LAN to build and update its table of source addresses. A table is maintained for each LAN connected to the bridge. The tables are updated when new packets are detected or when addresses expire from nonuse after a specified time. If a packet contains an address that has not yet been learned, it is sent out over all active links. The best path is determined by an industry-standard spanning tree algorithm, based on such factors as the number of hops from the designated root bridge and speed of the links. Any redundant paths are put in standby mode and used only in case of primary link failure. Source Routing Transparent The source routing transparent (SRT) bridging method combines Source routing and transparent bridging, permitting the data of both to be passed over the same network. With source routing transparent (SRT), the routing information field indicator is used to distinguish between frames using Source routing and frames using transparent bridging. Transparent bridges, including those supporting Ethernet, do not alter the routing indicator, but Source routing bridges do change the routing indicator, setting it to 1. By inspecting the routing indicator, the source routing transparent (SRT)-compliant device can determine whether the frame requires transparent bridging or Source routing (Exhibit 2). In supporting both bridging methods, Source Routing Transparent-compliant devices eliminate the need for multiple types of internetworking equipment and separate network facilities. Previous screen The SRT Bridge Is Capable of Both Source Routing and Transparent Bridging

Ramifications of SRT In several key areas, Source Routing Transparent enhances the capabilities of source routing and transparent bridging, providing compelling advantages to users. Route Discovery Transparent bridging employs a simple method of forwarding packets and can learn other locations based on newly encountered addresses. With Source routing, the originating device sends out explorer packets to discover the route between source and destination end stations. source routing transparent (SRT) not only supports both methods, it uses less route discovery overhead than Source routing because both the source and destination end stations can simultaneously discover the route. Congestion and Load Balancing Source routing does not automatically reroute around congested or failed links or perform load balancing. Transparent bridging does not make use of idle standby paths during normal operation or offer load balancing in the event of congestion. SRT makes more efficient use of the available bandwidth by switching to standby paths in case of a failure on the primary link. Network Consolidation SRT-compliant devices support both 4M-bps and 16M-bps Token Ring Lans and can consolidate Ethernet and Token Ring over shared WAN facilities. Platform for Interoperability Although source routing transparent (SRT) does not translate packets, its support of both Source routing and transparent bridging provides an internetworking platform and makes possible the use of applications that will perform such translations. SRT-compliant bridges do not solve the problem of consolidating System Network Architecture and Token Ring traffic over the same network. This is an important concern for many users who want to carry forward their considerable investments in fully functional System Network Architecture equipment, yet take advantage of Token Ring LANs for peer-to-peer networking. The optimal solution lies in choosing an source routing transparent (SRT)-compliant internet working device that also supports a capability known as SDLC passthrough. SDLC Passthrough Much of the installed base of SNA equipment has not been upgraded to support Token Ring connectivity, so the support of SDLC traffic is a significant feature of today's internetworking devices. Before SDLC support, expensive adapters were required to attach, for example, a 3174 cluster controller to a Token Ring LAN. Alternatively, the two traffic types had to be separately routed over the WAN. The latter solution entailed an overhead Previous screen burden and the cost of two separate networks. With internetworking devices that support SDLC traffic, users are finally able to consolidate parallel networks into a single multiprotocol backbone. Among the benefits of this arrangement is that it encompasses the SNA environment without any modification to the installed base of cluster controllers and front-end processors. Thus, SDLC passthrough is useful for integrating different generations of equipment. In combining SDLC serial traffic from SNA devices with LAN traffic—as well as supporting such commonly used non-SNA protocols as NetBIOS, TCP/IP, and IPX/SPX—communication costs can be greatly reduced. SDLC passthrough also provides an economical migration path from older SNA cluster controllers and front-end processors to Token Ring peer-to-peer networking. Cluster controller and Front-End Processor performance is improved by speeding up the communication lines from the slow 4.8K- and 9.6K-bps serial lines employed by SNA to WAN link speeds of 64K bps. Reliability is also improved because the SDLC data stream is encapsulated within the Token Ring frame format for routing across the WAN. This allows the advantageous use of redundant routes not available under SNA. If a line fails, traffic can be rerouted in an average of three seconds—well within the timeout threshold of SNA sessions. SDLC passthrough also allows a Permanent Virtual Circuit to be predefined between network end-points. Some vendors allow 16,000 unique virtual circuit numbers in either a point-to-point topology involving a front-end processor (FEP) and controller, or multidrop topology involving as many as 255 cluster controllers. Each connection has a station number as well as an assigned circuit number. Token Ring frames containing the encapsulated SDLC frames are routed to the appropriate network port. Network Management Selecting the right internetworking equipment vendor can actually facilitate network management. Not only can the vendor's management system monitor and control its own internetworking devices, but there may be enough flexibility to manage Token Rings, TCP/IP networks, and devices compatible with the Simple Network Management Protocol. Compatibility with IBM's LAN and host management systems—LAN Manager 2.0 and NetView—is another advantage of a good equipment choice. Of particular benefit to IBM users would be the network management system's built-in protocol analyzer features for troubleshooting the integrated network. Protocol analysis is not an inherent feature of IBM's network management systems, so users must typically spend $25,000 or more for a third-party solution. When protocol analysis is an integral feature of the internetworking equipment vendor's network management system, the cost is only $3,000 to $4,000. A Case Study Recently, one brokerage services firm implemented a plan to upgrade its network. A key component of the plan was to move from System Network Architecture to LANs to achieve greater efficiencies and economies in service delivery. The resulting network would have to serve multiple departments, including real estate, tax, ordering, unit trust, information systems development, futures trading, personal financial management, investment banking, and retail sales. Competitive Environment Previous screen Trading companies rely heavily on their networks to serve customers better and stay competitive. Consequently, more and more brokerage houses are foregoing the traditional SNA architecture, with its single-protocol low-speed analog lines and polled, multidrop modems. One byproduct of consolidation activity in the brokerage industry is the diversity of protocols that must be accommodated on the enterprise network. Another byproduct is that the redesign of corporate networks opens up the opportunity to employ more modern network protocols (e.g., Novell's NetWare). The integration of a variety of microcomputers and protocols (e.g., IBM's SNA/3270, Digital Equipment Corp.'s DECnet and Local Area Transport, and TCP/IP) gives brokerage firms the networking flexibility needed to meet the current and emerging demands of their customers. Evaluating the Products To achieve this level of integration, the products of many internetworking vendors are typically evaluated. Most brokerage houses with mixed-vendor environments disqualify IBM's source-routing bridges at the start because they cannot support non-IBM protocols. Furthermore, IBM bridges operate on a PS/2 platform and the presence of a keyboard invites tampering, representing a potential security problem. Traditional routers are also usually disqualified early in the evaluation process because they do not operate well enough in the SNA environment. A bridge represents a good choice in the multiprotocol networking environment precisely because it overcomes these problems and because it is economical and relatively easy to install. The Brokerage Network This New York brokerage house has more than three hundred offices worldwide connected through modems to the network at its headquarters in Manhattan. There, a star network links six home-office buildings (Exhibit 3). Each building has its own LANs ,which are interconnected over different media at the T1 rate (1.5M bps), at minimum. Because higher-rate private lines in large cities are so expensive, there are 10M-bps and 45M-bps (Ethernet and T3 speeds) microwave links between several hubs. The T3 facilities support dual 16M-bps Token Ring LANs and miscellaneous other traffic, for example, 3270 terminal traffic and voice. Gateways are used for translation of Novell's Internetwork Packet eXchange and IBM's NetBIOS protocols, among others, between the Token Rings and the SNA host.

The Brokerage's Backbone Network Then there is the need to pass information between Token Ring and Ethernet LANs. This is effectively handled by a bridge that operates independently of the protocols being used. Such bridges facilitate the interconnection of the two LAN types. This is no small task because there are more than 1,000 connections on Token Ring and Ethernet LANs at this brokerage house. Although some traditional routers also offer this capability, they do so at a premium price. This brokerage firm did not need the routing capability, so spending more for devices that would have been used primarily for their protocol conversion capabilities was not justified. The rule among brokerage firms is to protect their networks against the potential Previous screen loss of vital information. The New York brokerage firm implemented its disaster recovery plan in association with ComDisco (Foster City, CA). Every night, all data from the home- office locations is sent to a ComDisco backup facility in New Jersey. Within ten minutes of a computer failure on the company's internet, lost data can be retrieved from the backup site. Network Management The bridge's network management system can isolate problems across the distributed computing environment. It actually does a better job of troubleshooting in such an environment than host-based management systems, despite residing on a single workstation. From the management console, any bridge can be remotely configured, performance statistics gathered, and real-time alarms observed. Moreover, the enterprise network can be quickly diagnosed with a segment-by-segment connectivity test to determine whether a problem is in the network or application software. For example, a remote bridge can be instructed to test the integrity of the connection between itself and a questionable server. Then the bridge can test the connection between the management system and the remote bridge, completing the network test. Any segment of the network may be tested in this way without the need for extra equipment and without dispatching technicians to remote locations. These management features achieve greater network uptime, configuration flexibility and economies in equipment and personnel. Specifically, the brokerage firm can accommodate incremental growth, mitigate congestion, and better protect itself against potential disasters. The Optimal Solution Companies of all types and sizes are turning to LAN internets to obtain the operational efficiencies and economies they require to meet the informational needs of a diverse and demanding constituency that often includes consumers, shareholders, regulators, subsidiaries, and strategic partners. For users in mixed-vendor environments, with a large installed base of System Network Architecture equipment, the benefits of Source Routing Transparent-compliant internetworking devices that also support SDLC passthrough are clear. They provide access to and management of a broad range of computing resources from a single workstation or personal computer. And they provide organizations with the performance advantages and configuration flexibility needed to cope with the new economic and competitive realities—advantages that source route bridging and transparent bridging alone do not provide when information must be passed between diverse systems in multivendor environments. Author Biographies Nathan J. Muller Nathan J. Muller is an independent consultant in Oxford CT, and specializes in advanced technology marketing and education. He has 20 years of communications industry experience, writing extensively on many aspects of computers and communications, and publishing five books and more than 300 articles. He has held many technical and marketing positions with such companies as Control Data Corp., Planning Research Corp., Cable & Wireless Communications, ITT Telecom, and General DataComm, Inc. He has an Previous screen MA in Social and Organizational Behavior from George Washington University.