30 July 2003—Suppose that your company, like United Parcel Service of America Inc. (UPS, Atlanta, Ga.) and many other businesses, has a data network that represents a lot of investment, knowledge, and expertise. Suppose, further, that your company’s data network, like that of UPS, connects to and relies on a high-speed backbone network to transmit information cross-country or around the world. Although the two networks don’t speak the same language—that is, use the same telecommunications protocols—that’s O.K., because the capability to translate between the networks has long been established and works fine. Now suppose that you hear that the backbone network’s protocol is going to be replaced by a whole new one. With that news, you, or any business dependent on that backbone network, would be concerned that your data network would still be able to speak seamlessly with the new network protocol, whatever it is.
That’s why the demonstration of an interoperability standard between frame relay, the most common corporate data network protocol, and multi-protocol label switching (MPLS), the upstart backbone protocol, at the telecommunications convention, SuperComm, in Atlanta, last June has businesses breathing easier. Globally, businesses have installed a total of 12 million frame relay connections and, in 2002, spent US $14.5 billion on frame relay services, according to research firm Vertical Systems Group Inc. (Westwood, Mass.).
The interoperability announcement means that businesses won’t have to rip up their reliable, secure, trusted frame relay data networks and spend lots of money to replace them. What might seem like a rather obscure announcement has financial repercussions of billions of dollars. ”People want to make sure they can keep using frame relay for a long, long time,” says Rosalyn Roseboro, program director for switching and routing at research house RHK Inc. (South San Francisco, Calif.).
As a protocol, frame relay is a bit slow, with a typical maximum transmission speed of a T1 line (about 1.5 Mb/s). However, it is cheap, well known, highly reliable, and very flexible. Corporate frame relay networks also have one additional and unbeatable characteristic: they’re already built and paid for. Including training and equipment, frame relay technology represents ”a tremendous 12-year embedded investment of $20 billion worldwide,” says Roger Ruby, chair of an industry group called the Frame Relay Forumbefore it merged with the MPLS Forum to create the MPLS and Frame Relay Alliance (Fremont, Calif.) in April.
But corporate networks don’t stand alone. Instead, they use dedicated lines that stretch to the nearest telecommunications company node—a distance that may be less than a mile to a few miles. Once corporate lines reach a telecom node, they interface with the carrier’s high-speed backbone network. There, the data gets carried by, but not necessarily converted into, the backbone network’s protocol.
There are many kinds of backbone networks, and they use different protocols. The specifics for each network depend upon its purpose, when it was built, the state of technology at the time, and many other factors. However, frame relay networks commonly link up to backbone networks that use a protocol called asynchronous transfer mode (ATM). ATM is common primarily because it was extensively deployed in the mid-1990s as the sole broadband backbone technology capable of carrying voice, data, and video. Interoperability between frame relay and ATM was established in 1995, and the two have worked together ever since.
Despite that, ATM is not necessarily the perfect telecom mate for frame relay. Both are packet-type technologies, but whereas frame relay uses a variable-length data packet (a frame) of up to 4096 bytes, ATM uses a fixed-length data packet (a cell) of 48 bytes (plus a 5-byte header). ”Whenever frame relay is carried over ATM, it has to go through a segmentation process to split the frames into cells, and then the frames have to be reassembled at the far end of the network,” explains Andrew Malis, chairman and president of the MPLS and Frame RelayAlliance.
Segmentation, however, is not perfect. Not all cells are completely filled with the frame’s data, so extra cells are produced, lengthening the message. In addition, every one of the many ATM cells built from a single frame gets a new header. These translation steps add overhead. ”On average,” Malis says, ”there is a 15�20 percent added overhead. This takes additional capacity in the network.” To send a 100-byte frame relay message, for instance, an ATM network must carry 115�120 bytes.
This is where MPLS comes in as a replacement for ATM. Like frame relay, MPLS uses a variable-length data packet, eliminating the need to chop big frames up into tiny cells. There’s no need to add headers, either. In effect, an MPLS network need carry only 100 bytes to send a 100-byte frame relay message.
Besides the lower data overhead, MPLS has another advantage over ATM: it’s faster. ”The general opinion is that very high bit rates are harder to do on ATM than on MPLS,” says Rick Townsend, president of the ATM Forum and chair of the forum’s technical committee. He adds that OC-192 (9.9 Gb/s ) is ”about the transition point where MPLS starts to look better than ATM.” Currently, OC-192 is the top speed for most optical fiber backbone networks, but OC-768 (40 Gb/s) is on its way.
No matter how much better MPLS may be than ATM as a backbone technology, ”MPLS can’t replace ATM unless it can carry frame relay,” says John Killeen, director of global network services at UPS. And now it will. MPLS-frame relay interoperability means that UPS and every other business with a frame network can keep it running indefinitely.
The next step by the MPLS-Frame Relay Alliance will be to work with the ATM Forum to establish interoperability standards between ATM and MPLS. That interoperability is important because MPLS will be implemented over a period of years. At first there will be islands of MPLS technology in a sea of ATM, and then the reverse is expected. Ultimately, many in the industry believe that ATM will vanish from the backbone completely.