The debate about frame relay vs. ATM as WAN alternatives was about packets vs. cells.

Frame relay, the outgrowth of slower, more careful, error-correcting X.25, is packet technology designed to carry variable-length frames over high-quality connections such as fiber, which was just coming into its own in the early 1990s when frame relay started its heyday.

ATM was designed in the 1980s to deliver five distinct levels of QoS, so users could send traffic with greater or less delay. ATM employed regular-sized packets called cells. If the payload was longer than the 48 bytes available, it would be split up and packaged in other cells. This was not true of frame relay, which accommodated frames of widely varying length.

Key to ATM’s charm was that it could emulate direct circuits and guarantee bandwidth, a shortcoming of frame relay. The latter could offer more bandwidth than customers ordered when their traffic load increased, but there were no guarantees that the extra capacity would be available from the network.

One downside of ATM was the cell tax - the fact that for every 48 octets of payload a cell required five octets of header - about 10% of the total number of bytes that made up a cell.

ATM didn’t do well initially because it was only offered as a service on T-3 connections, whose 45Mbps capacity was more than most businesses needed or wanted to pay for. Frame relay, on the other hand, was less expensive than the common alternative - dedicated circuits -- and was cost-effective at 56Kbps.

Frame relay won in the WAN. ATM lived on, though in carrier core networks, where it is slowly being decommissioned. Ultimately, both frame relay and ATM are losing as they are being pushed out by MPLS.

< Return to the list of arguments >