Frame Relay

Frame Relay

Frame relay is a byproduct of the ISDN standards work. Frame relay was initially expected to be offered on Broadband ISDN networks, but it is currently marketed as an upgrade to X.25 packet-switched networks.

Frame relay makes sense if your enterprise network interconnects more than four sites and network traffic is fairly heavy. If your network traffic is primarily from terminals, then X.25 is a better service option; however, if your network traffic is from workstations and PCs, then frame relay is probably your best bet.

Backtrack To X.25

Packet-switched networks are suitable for connecting geographically dispersed locations, because instead of directly connecting each remote site to the headquarters or connecting each site to every other one, packet switching provides a "cloud" network into which sites can connect. Once attached to the cloud, any site can communicate with any other. Each site has a local connection, which is then connected into a larger packet-switched network.

Packet-switched technologies maintain virtual connections between two users. It appears to the end users that they are directly connected, when, in reality, the connection may go through intermediate points. Virtual connections can either be permanent or temporary. A permanent virtual circuit is guaranteed to be available at all times. With a temporary, or switched virtual circuit, the connection is set up for the duration of the transmission, then broken down.

X.25 is a packet-switched technology that has been used since the early 1970s. It was designed to transport character data from host terminals, and for such low-bandwidth applications, X.25 delivers adequate performance. But stuffing Ethernet or Token Ring traffic through a 56/64Kbit/sec X.25 virtual circuit is much more difficult.

X.25 was developed when the telephone system was dominated by copper lines, and it necessarily included a large number of error-detecting and -correcting routines to handle the effects of noise, dropped calls, and other problems. However, since the vast majority of the U.S. phone network is fiber-optic, the error-detecting and -correcting code in X.25 can be unnecessary overhead.

And so frame relay was developed—originally to run in the ISDN "D" channel. Frame relay retains many of X.25's good traits, including packet switching, while dropping other traits, such as the overhead. Frame relay was developed under the aegis of the CCITT I.122 and the corresponding American National Standards Institute committee 1.606.

Jump To Frame Relay

Frame relay's benefits include low overhead, high capacity with low delay, and reliable data transfer over existing public networks. It is largely designed to be a public service for interconnecting private local area networks, although private networks can be built.

Frame relay is generally considered to be able to provide service at speeds up to 2.048Mbits/sec, although some service providers claim that they will be able to squeeze out 45Mbits/sec. Public frame relay carriers have rolled out frame relay services in the increments of 56Kbits/sec, fractional T1, and full T1. Frame relay services will come from the RBOCs, interexchange carriers, and value-added networks (VANs), including British Telecom, Pacific Bell, CompuServe, and Wiltel.

Frame relay operates on OSI Layers 1 and 2, whereas X.25 operates on Layers 1, 2, and 3. Because frame relay operates under the network layer, it is independent of the upper-layer protocols, such as TCP/IP or IPX. This arrangement delivers greater flexibility.

Frame relay achieves high throughput with low delay by eliminating the overhead of error detection and correction. Data integrity is insured only through a cyclic redundancy check, and any corrupted packet is discarded. When necessary, the network-layer device—usually a router—handles retransmissions. That device must detect the transmission failure and request a retransmission. Although errors will occur less frequently because of the more reliable phone lines, the responsibility of handling them has simply moved to another device that must be capable of handling the added work. Frame relay will also drop packets when the network is too congested.

Frame relay offers some pricing advantages over other WAN technologies, which is priced on a distance- and capacity basis. T1, for example, is priced according to the distance between two sites; the greater the distance, the more expensive the T1 line. X.25 is priced based on usage; the higher the volume of traffic, the more you pay.

X.25 is used rather infrequently in the United States because leased lines are plentiful and relatively inexpensive; however, X.25 is widely used internationally. X.25 is often the only reliable WAN method, as many countries' phone systems are outdated and leased lines are not readily available. As a follow-on to X.25, frame relay will be popular in Europe, as well as for connecting U.S. sites to those abroad.

Frame Relay And You

Like LANs, frame relay uses a variable length packet, with sizes ranging from a few bytes to more than 4,000 bytes. A variable packet length enables it to accommodate the LAN's bursts in traffic; however, because the delay is unpredictable, frame relay is not especially suitable for voice or video traffic—only data.

Each frame includes an 11-bit address field, the Data Link Connection Identifier (DLCI), which supplies the virtual circuit number that corresponds to a particular port on a switch. Frame relay offers independent packet addressing, which also reduces overhead. Private virtual circuits may be set up between addresses.

A variety of devices can be used to bring frame relay to your LAN. If you choose a public frame relay service, you'll need frame relay equipment on your premises, as well as a connection to the frame relay network via the local loop.

Most corporations use a router with a frame relay interface card. These routers may use specialized hardware or run on a PC.

You can also use a multiplexer, commonly referred to as a mux, that accepts data from the LAN and forwards it to the frame relay network. A mux is handy in a point-to-point configuration, but if your network uses multiple protocols or has a complex configuration, then you should use routers.

If you implement a private frame relay network because of security, reliability, or control issues, you will need to purchase frame relay switches. The added benefits may offset the costs of investing in your own switching equipment.

Frame relay equipment has the advantage of being relatively similar to X.25 equipment. In many cases, the manufacturer does little more than change the software to give an X.25 device frame relay functionality. This similarity can make upgrading from X.25 to frame relay more cost-effective than switching to another WAN technology.

Frame Relay Foibles

Frame relay comes in two versions: Frame relay 1 sets up a permanent virtual circuit, in which the connection must be maintained at all times. One of the touted benefits is that frame relay offers "bandwidth on demand." This benefit should not only mean that you can dial up extra bits per second when the LAN traffic surges, but also that you can dial up the service itself as needed.

The next version of the standard, frame relay 2, will offer this switched virtual circuit service. When switched virtual circuits do become available, frame relay will become more flexible, since users can dial up the service.

Frame relay does not handle congestion with grace. When there's too much traffic on the frame relay network, the frame relay equipment drops frames. Because network traffic travels in bursts, the receiving device's buffer can easily become overloaded. The device discards the packets and waits for the end nodes to notice and retransmit. This strategy is not reliable for timing-dependent networks.

The two methods for handling congestion within frame relay are Explicit Congestion Notification (ECN) and Consolidated Link-Layer Management (CLLM). Neither method was initially standardized, and it remained the prerogative of each vendor and service provider to implement it differently or not at all. This lack of conformity makes communication among different vendors' frame relay products difficult at best. ECN offers the ability to communicate traffic overload problems downstream and upstream. ECN can be done from the source node in the direction of the data flow, which is called Forward ECN. Or it may be done from the data flow back toward the sending node. This process defines Backward ECN. The congestion flag is set in the DLCI address, so that all sending and receiving nodes know that the network is congested. With CLLM, one of the DLCIs is used to send link-layer control messages. These messages can be coded to include the type of congestion and the addresses of the DLCIs involved.

In addition to ECN and CLLM, the router can handle the congestion, although errors can be handled more quickly at the lower layers of the OSI stack.

Each frame relay carrier offers a different set of access rates, pricing, and price discounts. When selecting a frame relay carrier, check the link speed and the committed information rate, or the data rate between two sites, that the carrier guarantees. You should specify a committed information rate that's below the link speed you think you'll need; this way the frame relay network can accommodate bursts of LAN traffic. In some instances, you'll have to lease a line between your site and the frame relay point of presence.

What To Do, What To Do

Frame relay is a good interim media. Technically, it's better for LAN traffic than is X.25, and it has particular applicability for international networks. Yet its ultimate success will be determined by factors beyond technology. If the service providers price frame relay aggressively, then users will be encouraged to try a new service. Frame relay's success may be dampened by technologies such as ATM and SMDS that not only offer high speeds but, more importantly, are able to carry voice and video as well as data. For it is in the high-speed integration of data and voice that a true enterprise LAN/WAN can be constructed.

This tutorial, number 54, by Patricia Schnaidt, was originally published in the February 1993 issue of LAN Magazine/Network Magazine.