Frame relay gets the call >BY NINA SABERI

Data networking is changing. Increased competition among frame relay service providers is driving prices down and stimulating increased demand from users seeking more economical and efficient connectivity for large branch networks. To counter the margin pressure from price erosion, public network carriers are aggressively pursuing opportunities to reduce costs and provide value-added services. A new generation of frame relay access devices will allow both carriers and their customers to achieve their objectives.

From a user's standpoint, the growth and acceptance of frame relay is a result of the price differentials that arise from the inherent cost-effectiveness of frame relay technology. For a large branch network, frame relay offers a 25% to 30% price advantage over leased lines, as well as the ability to temporarily provide peak usage bandwidth beyond normal usage.

Analysts expect that half the installed base of long-haul, interexchange circuits will be converted to frame relay service over the next five years. More than 70% of these leased circuits are carrying IBM's Systems Network Architecture (SNA) protocol. Converting these circuits to frame relay is a major market opportunity for carriers, and success in the SNA-over-frame relay segment will play a key role in determining long-term market penetration. Figure 1 shows a complete SNA-over-frame relay solution.

Basic Economics As a starting point for economic analysis, it is helpful to divide the SNA opportunity into SNA-only and SNA-plus-local area network requirements. If a user has a large, multibranch SNA-only network, replacing a dedicated, leased-line backbone with a shared frame relay backbone can mean substantial savings in line cost. But a customer must also factor in the price of the frame relay access device (FRAD) at every remote site, including the cost of installing, maintaining and managing that remote equipment.

In many cases, the cost does not justify the savings. The most attractive solution for a large part of the SNA-only network opportunity is for carriers to provide the conversion from SNA to frame relay within the network infrastructure itself. This solution not only eliminates access device-related costs for users, it also provides carriers with added value and a substantial differential at a low incremental provision cost per SNA site. A well-structured service based on this approach represents a compelling economic case for migration.

For customers who have both SNA and LAN traffic, the economic model is entirely different. Not only is there an inherent lower provision cost in frame relay than in the leased line that it replaces, but because of frame relay's ability to transport both SNA and LAN traffic, going from two or more access lines to a single shared access line and eliminating the administrative overhead of operating two parallel networks also generates cost savings.

Moreover, remote sites with LANs always need an access device to provide the routing function. These factors more than offset the cost of the FRAD, and the economic case for migration is extremely strong. The critical issue becomes the ability of the integrated access solution to ensure the same service levels for both SNA and LAN networks as the alternative leased-line networks.

In reality, many real-world networks are a mix of the SNA-only and SNA-plus-LAN models, and to win the battle for SNA-over-frame relay market share, a carrier must excel in both areas. It is also important to realize that many organizations lack the internal resources or focus to design and roll out a network infrastructure by themselves. Carriers can respond by offering a total value-added service with frame relay capability in the network for both SNA-only and SNA-plus-LAN models and by managing the customer premises equipment and frame relay access equipment from the carrier's network operations center.

Challenges Although most of the economic arguments seem to favor frame relay, carriers must address some important technological considerations in implementing their services. One of prospective users' major concerns is the ability to maintain service levels for critical SNA applications.

SNA was originally designed for the leased-lined infrastructure. Unlike leased lines, however, frame relay services are shared among multiple users and multiple traffic types. While frame relay offers more bandwidth for the same money on average, the concern is that the bandwidth available at any particular time is not guaranteed. SNA traffic is typically transaction-oriented, with users being extremely sensitive to the delay between hitting an action key and getting a response. Furthermore, inconsistency in delivery or response times can cause SNA sessions between the branch devices and the mainframe to become prematurely terminated.

Despite several years of promises, the challenges to provide frame relay access for SNA traffic have not been overcome by routers. Even the current solution, data link switching, only provides SNA support at a high cost for bandwidth overhead, processing overhead and scalability and management restrictions. The solution is the latest generation of FRADs.

The new FRADs for the customer premises are specifically designed to meet the needs of consolidating legacy SNA traffic with LAN-over-frame relay. While these devices can act as a simple FRAD to accommodate just SNA traffic, the majority market and economic incentive is to consolidate the tail circuits from multiple network protocols. This approach encompasses both FRAD and router technology in a single device.

Some of the considerations for an SNA FRAD are its support for the wide variety of SNA implementations and permutations in use, its ability to minimize SNA response times and maintain SNA session integrity, and its support for IBM management, including NetView and LPDA. For SNA-plus-LAN applications, the FRAD must provide high-performance routing for LAN traffic and provide sophisticated bandwidth allocation and congestion control mechanisms to ensure that the needs of critical SNA traffic are always met.

Providing multiprotocol frame relay access for a branch site places demands on the access devices that are much higher than those placed on access routers or simple FRADs. Independent testing has shown that even at the typical branch access speed of 56 kb/s, actual data throughput achieved by multiprotocol FRADs can differ by a factor of two.

SNA response times can vary by at least this amount. For data center access and-even more so for network-based central office deployment-the ability to fully use multiple T-1 access lines and the logical capacity to support hundreds (or thousands) of remote sites is vital. For CO systems, fault-tolerant platforms and channelized T-1 access are critical requirements.

As frame relay evolves, demand for access speeds above 56 kb/s is likely to increase, and more sophisticated implementations will use priority PVC schemes, large virtual circuit meshes, switched virtual circuits and frame relay-to-ATM internetworking. With the rapid pace of development in networking, an access device's potential to evolve with market demands is more important than ever.

SNA networks represent one of the largest opportunities for carriers seeking to build-or protect-market share. Frame relay-based services can present a compelling economic case for SNA networks to migrate, and those carriers with the solutions that best reflect the technical and economic needs of SNA users will be the winners of a key battle in the market share wars.

Nina Saberi is President and Chief Operating Officer for Netlink Inc., Framingham, Mass.