More benefits of oversubscription

In a recent column we said oversubscription of a frame relay port was a good idea. This discussion generated so many questions that it seems worthwhile to delve a bit deeper into the subject.

The fundamental idea of oversubscription is that you assign committed information rates (CIR) for the permanent virtual circuits (PVC) associated with a port such that the sum of the CIRs may indeed exceed the port speed. This guarantees that the network bandwidth is available if and when you need it. (The operative word in the last sentence is "network." Obviously, if you oversubscribe the port with CIRs that total more than the speed of the port, the port will be the potential bottleneck in the equation.)

The fundamental goal is to safely conserve on expensive port costs. The intermittent, bursty nature of frame relay traffic virtually guarantees that all the PVCs are not going to require maximum CIRs at the same time. The telephone company has been playing this same type of game for decades by oversubscribing users to the number of phone switch ports available.

This whole idea of oversubscription is deeply rooted in the fundamentals of statistical multiplexing. The underlying reason that you save money by using statistical multiplexing is that data traffic tends to be quite bursty.

With time division multiplexing technology, each individual channel is permanently assigned a portion of the bandwidth. For instance, thinking back to the days of asynchronous terminals, eight terminals at 1200 bit/sec could share a 9600 bit/sec circuit (8x1200=9600). OK, so we're ignoring the overhead, but you get the general idea.

With statistical multiplexing, the multiplexer has the intelligence to assign bandwidth on the circuit as needed. For the above situation, one would typically set each of the eight terminals to 4800 bit/sec, resulting in a four-to-one oversubscription. The odds of all the terminals being active simultaneously are small enough that one would typically never see performance degradation. In the event that all of the terminals were active simultaneously, the buffers in the multiplexers would store the data temporarily until bandwidth became available. (Again, we're ignoring the overhead factor, but typically the statistical multiplexing saves so much bandwidth that the higher overhead doesn't have a major impact.)

This same type of oversubscription takes two forms in regards to frame relay services. First, frame relay carriers can "play the odds" (a.k.a. traffic engineering) so that they may provide service to a large number of customers over shared rather than dedicated bandwidth. Second, and this is the one we're concentrating on, frame relay customers can oversubscribe a port so that they can have guaranteed bandwidth through the carrier network to a number of points such that the sum of the guaranteed network bandwidth might significantly exceed the port speed.

---

Steven Taylor, consultant and broadband packet evangelist, and Joanie Wexler, an independent networking technology editor and writer, team up to bring you this analysis and commentary. Taylor specializes in education and market analysis, and Wexler adds incisive reporting and research. For more detailed information on most of the topics discussed in this newsletter, connect to www.webtorials.com, the first Web site dedicated exclusively to market studies and technology tutorials in the Broadband Packet areas of Frame Relay, ATM, and IP. Feedback and additional topic ideas are welcome. Please contact taylor@webtorials.com or joanie_wexler@mindspring.com>

Frame relay at the crossroads
Network World, 05/31/99

Suppressing silence saves money
Network World, 02/01/99

Archive of Network World on Frame Relay newsletters