By Paul K. Wickre
Network World, 6/23/97

Domestic frame relay usage is expected to more than double in the next two years. Usage will expand from 160,000 interexchange carrier (IXC) ports in service at the end of 1996 to more than 400,000 by the end of 1998, a situation that threatens to make carrier backbones look like major highways on a summer holiday weekend.

Increased traffic is straining the ports now in service, while provisioning delays mean the queue of new customers waiting to obtain connections continues to grow. In short, carriers are having a hard time keeping up with demand, so you'd be wise to anticipate demand-related problems and come up with contingency plans.

"We are seeing increases of traffic levels measured in terabytes in the 15% range month-to-month,'' says Melanie Hanssen, senior manager of frame relay marketing at MCI Communications Corp. in Irvine, Texas. LDDS WorldCom and Sprint Corp. confirm that their traffic levels also are growing by 15% to 20% per month.

This growth is primarily off the existing installed base of frame relay users; it will be compounded by the migration of SNA private lines to frame relay services. IXCs' frame revenues in the U.S. are currently in the $1 billion range, while private-line revenues eligible for frame relay conversion total about $4.9 billion.

"The 1997 to 1998 frame relay scenario will look like the snake that swallowed the football. Whether or not the carriers can digest the load and Fortune 1,000 private-line demand while maintaining service quality remains to be seen,'' says Fred McClimans, CEO of Current Analysis, a consultancy in Ashburne, Va.

Roadblocks ahead

But just wait - the massive growth driven by Internet-based multimedia applications, increased LAN traffic and the mass conversion of large, branch SNA shops will leave carriers with their hands full.

Frame relay carriers today offer service-level guarantees based on several factors including network delay. The major carriers generally offer one-way network response times of less than 250 msec. For example, Sprint guarantees 90 msec for a frame relay T-1 line, 105 msec for 256K bit/sec service and 115 msec for 56K bit/sec service. MCI has pushed the envelope further by guaranteeing 70 msec within the network cloud, not counting serialization delay from the customer premises equipment (CPE) device.

These figures are well within the timing tolerance of transaction-processing systems.

The key is that network delay is a function of traffic load. As traffic volume increases at peak times, so will delays. These delays may cause dropped sessions and lost data.

Network delays and transit times aren't the only concerns. By its nature as a fast packet service, frame relay lacks error correction. Frame relay assumes the higher layers within the application protocol, such as Synchronous Data Link Control for SNA, will perform the error recovery routines. Unfortunately, older protocols still in use, such as asynchronous and bisynchronous, don't have this capability.

The widespread use of poll spoofing techniques may also strip SDLC away, leaving SNA packets vulnerable. Even worse, frame relay allows for packets to be marked discard-eligible during periods of peak usage, bursting or severe congestion caused by internal network failures such as line cuts. If the CPE device is transmitting multiprotocol traffic above the committed information rate and the discard eligibility bit is turned on, there's a potential for serious data loss.

Today, carriers compete on price and service variables, but in the future they will probably differentiate their classes of service offered. Expect pricing structures to evolve based on quality and class of service.

How carriers will address service options, pricing and value-added features is still in the early planning stages. However, several carriers recently announced priority levels for permanent virtual circuits (PVC). For instance, MCI, Intermedia Communi-cations, Inc. (ICI) and LDDS WorldCom plan to roll out several levels of PVCs. The highest priority would be reserved for delay-sensitive traffic such as SNA or voice, the next level would be for mission-critical IP and the following levels would trail off in importance and quality.

Although this is a step in the right direction, priority PVCs have the same flaw as reservation schemes. If the carriers don't impose a premium pricing mechanism on superior service, all users will reserve bandwidth or claim that all of their traffic should be in priority PVCs across the backbone.

Think of it this way: If all travelers could fly first class at no additional cost, business, tourist and coach passengers would all clamor for first-class amenities. Price differentials force customers to make a rational, cost-benefit justification. MCI has plans to charge a ''small differential'' for the priority PVC, and Sprint also may vary pricing for its application-based class of service, scheduled for launch by year-end.

Steering clear of potholes

For the utmost degree of flexibility, consider frame relay-to-ATM interworking offers that carriers are now rolling out, particularly if your traffic mix will include data, voice and video.

As a safeguard against finger-pointing, have your carrier bundle intelligent CPE into a managed services offering. Second generation frame relay access devices and routers are increasingly sophisticated in prioritization schemes and proprietary error recovery from remote branch to host. In addition, incorporating the access layer, tail circuit, transport and remote termination simply makes good technical and business sense. More than 76% of Fortune 1,000 companies prefer this approach, according to Frame Relay Systems and Technology's recent survey.

Managed service also lets the carrier provide end-to-end network tuning, and supplies the tools needed for rapid fault diagnosis and service restoral. For example, vendors are increasingly incorporating a Remote Monitoring (RMON) probe, remote trace capability or software diagnostic product within the CPE. Visual Networks, Inc. of Rockville, Md., Sync Research of Irvine, Calif., and Concord Communications, Inc. of Westborough, Mass., offer a variety of products for remote diagnostics, troubleshooting and circuit management. Although these features increase the initial cost of deployment, they can deliver a huge return on investment by speeding service restoral to curtail lost revenue.

The managed service agreement should include performance and configuration management. The carrier will typically provide tools such as a customer-site network-performance monitoring workstation.

After selecting a network design, PVC prioritization, CPE choices and managed service, back up the whole frame relay network with an SLA that offers credits for downtime. This is a contractual relationship that ensures monthly statistical quality of service (QoS) for your entire enterprise. Negotiate this in conjunction with terms of service and volume of business and enforce it by using per-site reports of outages and performance.

Finally, set aside a reserve budget of 5% to 15% for anticipated priority PVC surcharges in case performance issues and congestion problems arise next year. Forewarned is forearmed.

Frame delay, frame loss ratio and response time measurements will begin to deteriorate and impact applications. Frame relay was never designed for the mix of traffic that now runs on the carriers' net. The notion of QoS comes from ATM, where constant bit rate and variable bit rate could be specified on an individual connection-oriented basis. Because most carriers plan to use ATM in the backbone and frame relay for the access layer, extending ATM QoS out through frame relay nets remains an architectural issue. Priority PVCs are a start, but more sophisticated schemes need to emerge, fully mapping ATM QoS provisions into delivered frame relay classes of service.

Although the ATM and Frame Relay Forums have debated this and issued standards - Data Exchange Interface and Frame-based User Net-work Interface - re-quests for comment that address mapping service quality from ATM to frame relay have yet to be commercially adopted. In the long run, carriers will offer proprietary approaches that will approximate what the standards bodies have tried to implement.

Understanding the growth scenario ahead, users should proactively work with their carrier partners to examine or implement these issues. With a bit of planning and foresight, frame relay users will continue to receive superior price performance for their geographically dispersed branch networks without unforeseen headaches.