10.2 Frame Relay

---

Frame relay is a stripped-down version of the X.25 protocol for packet-data networking, which was designed to transport data very reliably over noisy analog-line networks. In running over cleaner digital-line networks, however, frame relay eliminates many of the functions of X.25, including node-to-node error checking and correction. In stripping away unnecessary functions and relegating error correction to the CPE at the edge of the network, frame relay could transport data at much higher speeds. Whereas the speed of X.25 topped out at 56 Kbps over analog facilities, frame relay offers up to 45 Mbps over more reliable T-carrier facilities, and even higher over optical carrier (OC) facilities, making it better suited for interconnecting LANs at native speeds over a WAN.

With LANs becoming popular in the 1980s, there was a growing need to interconnect them over the WAN. Point-to-point T-carrier lines were becoming commercially available, but they were cost prohibitive for all but the largest companies. Frame relay was developed specifically to provide LAN interconnectivity as a carrier-provided service, eliminating costly leased-line charges based on the distance between corporate locations.

With the increasing use of digital facilities, there is less need for error protection of the kind offered by X.25. At the same time, the end devices increased in intelligence, processing power, and storage, making them better at handling error control and diverse protocols. Consequently, the communications protocols used over the network may be scaled down to their bare essentials to greatly increase throughput. This is the idea behind frame relay, which can support voice traffic, as well as data, packaged in variable-sized frames of up to 4,000 bytes in length.

Frame relay was introduced commercially in May 1992. It initially gained acceptance as a method for providing end users with a solution for data connectivity requirements, such as LAN-to-LAN connections. Frame relay provided both an efficient and flexible data transport mechanism and also allowed for a cheaper bandwidth cost associated with connecting legacy SNA networks.

Whereas X.25 operates at the bottom three layers of the OSI reference model, frame relay operates at the first layer and the lower half of the second layer. This cuts the amount of processing by as much as 50%, improving network throughput. Although the frame relay network can detect errors, it does not correct them. Bad frames are simply discarded. When the receiving device detects corrupt or missing frames, it can request a retransmission from the originating device, whereupon the appropriate frames are sent again.

10.2.1 Advantages of Frame Relay

The most compelling advantages of a carrier-provided frame relay service include the following:

• Improved throughput/low delay: Frame relay service uses high-quality digital circuits end to end, making it possible to eliminate the multiple levels of error checking and error control. The result is higher throughput and less delay compared to legacy packet-switched networks like X.25.

• Any-to-any connectivity: Any node connected to the frame relay service can communicate with any other node via point-to-point PVCs or dynamically via switched virtual circuits (SVCs).

• No long-distance charges: Since frame relay is offered as a service over a shared network, the need for a highly meshed private-line network is eliminated for substantial cost savings. There are no distance-sensitive charges with frame relay, as there are with private lines.

• Oversubscription: Multiple PVCs can share one access link, the aggregate bandwidth of which can even exceed the port speed of the frame relay switch. In oversubscribing the port, multiple users can access the frame relay network on a contention bases, eliminating the cost of multiple private-line circuits and their associated CPE for further cost savings.

• Higher speeds: Whereas X.25 offers up to 56 Kbps, frame relay service supports transmission speeds up to 44.736 Mbps. If the frame relay switches in the network support Frame Relay Forum Implementation Agreement 14 (FRF 14), speeds at the OC-3 rate of 155 Mbps and the OC-12 rate of 622 Mbps over fiber backbones are possible.

• Simplified network management: With fewer access circuits and less equipment to monitor, companies subscribing to a frame relay service can reduce their management requirements. And under a managed service, the carrier provides proactive monitoring of CPE and network maintenance on a 24/7 basis, assuming the entire management burden from the customer, even if a firewall is implemented by the routers operating system.

• Intercarrier connectivity: Frame relay service is compatible between the networks of various carriers, through network-to-network interfaces (NNIs), enabling data to reach remote corporate locations that may not be served by the primary carrier.

• Performance reports: Customers can manage their frame relay service to maximum advantage. Accessible on the carrier’s secure Web site, network reports are available for utilization, errors, health, trending, and exceptions.

• Service-level guarantees: Frame relay service providers offer customers SLAs that specify availability as a percentage of uptime, round-trip delay expressed in milliseconds, and throughput in terms of the committed information rate (CIR). If the carrier cannot meet the SLA for a given length of time, it credits the customer’s invoice accordingly.

10.2.2 Types of Circuits

The two primary types of virtual circuits supported by frame relay are SVCs and PVCs. SVCs are analogous to dial-up connections, which require path setup and tear down. A key advantage of SVCs is that they permit any-to-any connectivity between devices connected to the frame relay network. In contrast, PVCs are more like dedicated private lines; once set up, the predefined logical connections between two sites stay in place. Another type of virtual circuit is the multicast virtual circuit (MVC), which is used to broadcast the same data to a group of users over a reserved data link connection in the frame relay network. This type of virtual circuit might be useful for expediting communications among members of a single workgroup dispersed over multiple locations or to facilitate interdepartmental collaboration on a major project. It can also be used for broadcast faxing, news feeds, and “push” applications from the corporate headquarters to branch locations.

The same frame relay interface can be used to set up SVCs, PVCs and MVCs. All three may share the same access facility. The behavior of the virtual circuits is programmed into the frame relay switches within the network. In supporting multiple types of virtual circuits, frame relay networks provide a high degree of configuration flexibility, as well as more efficient utilization of the available bandwidth.

The virtual circuits have a CIR, which is the minimum amount of bandwidth the carrier agrees to provide for each virtual circuit. If some users are not accessing the frame relay network at any given time, extra bandwidth becomes available to users who are on-line. The CIR of their virtual circuit can burst up to the full port speed. As other users come on-line, however, the virtual circuits that are bursting beyond their CIR must back down to their assigned throughput rates.

10.2.3 Congestion Control

Real-time congestion control must accomplish the following critical objectives in a frame relay network:

• Maintain high throughput by minimizing time-outs and out-of-sequence frame deliveries;

• Prevent session disconnects, unless required for congestion control;

• Protect against unfair users who attempt to hog the available network resources by exceeding their CIR or established burst size;

• Prevent the spread of congestion to other parts of the network;

• Provide delays consistent with application requirements and service objectives.

In the frame relay network, congestion can be avoided through control mechanisms that provide backward explicit congestion notification (BECN) and forward explicit congestion notification (FECN).

BECN is indicated by a bit set in the data frame by the network to notify the user’s equipment that congestion avoidance procedures should be initiated for traffic in the opposite direction of the received frame. FECN is indicated by a bit set in the data frame by the network to notify the user that congestion avoidance procedures should be initiated for traffic in the direction of the received frame. Upon receiving either indication, the end-point (i.e., bridge, router, or other internetworking device) takes appropriate action to ease congestion.

The response to congestion notification is dependent on the protocols and flow-control mechanism employed by the end-point. The BECN bit would typically be used by protocols capable of controlling traffic flow at the source. The FECN bit would typically be used by protocols implementing flow control at the destination.

Upon receipt of a frame with the BECN bit set, the end-point must reduce its offered rate to the CIR for that frame relay connection. If consecutive data frames are received with the BECN bit set, the end-point must reduce its rate to the next “step” rate below the current offered rate. The step rates are 0.675, 0.50, and 0.25 of the current rate. After the end-point has reduced its offered rate in response to receipt of BECN, it may increase its rate by a factor of 0.125 times the current rate after receiving two consecutive frames with the BECN bit clear.

If the end-point does not respond to the congestion notification, or the user’s data flow into the network is not significantly reduced as a result of the response to the congestion notification, or an end-point is experiencing a problem that exacerbates the congestion problem, the network switches implement congestion recovery procedures. These procedures include discarding frames, in which case the end-to-end protocols employed by the end-points are responsible for detecting and requesting the retransmission of missing frames.

Frame discard can be done on a priority basis; that is, a decision is made on whether certain frames should be discarded in preference to other frames in a congestion situation based on predetermined criteria. Frames are discarded based on their “discard eligibility” setting of 1 or 0, as specified in the data frame. A setting of 1 indicates that the frame should be discarded during congestion, while a setting of 0 indicates that the frame should not be discarded unless there are no alternatives.

The discard eligibility may be determined in several ways. The user can declare whether the frames are eligible for discard by setting the discard eligibility bit in the data frame to 1. Or, the network access interface may be configured to set the discard eligibility bit to 1 when the user’s data has exceeded the CIR; in which case, the data is considered excess and subject to discard. For users who subscribe to CIR=0, which moves data through the frame relay network on a best-efforts basis subject to bandwidth availability, all traffic is discard eligible.

10.2.4 Frame Relay Charges

Frame relay charges differ by carrier and may differ further by configuration. Accordingly, frame relay service charges may include the following:

• Port charge for access to the nearest frame relay switch, which is applied to every user location attached to the frame relay network.

• Local loop charge, which is the monthly cost of the facility providing access to the frame relay network. This charge may not apply if the customer’s building is directly connected to the carrier’s metro fiber ring, in which case the customer is charged only a one-time setup fee.

• Charges for the PVCs and SVCs, which are determined according to the CIR assigned to each virtual circuit.

• Burst capability, usually determined by the burst excess size. Most carriers do not specifically charge customers for bursting beyond the CIR, since this is one of the key advantages of going with the service in the first place.

• CPE, which includes the frame relay or internetworking access equipment optionally leased from the service provider and bundled into the cost of the service.

• IntraLATA/interLATA service charges. Usually, there is one price for “local” frame relay service and another price for “national” frame relay service. National frame relay service is usually defined as one that crosses LATA boundaries and involves the services of another carrier. Neither intraLATA or interLATA service is distance-sensitive with regard to pricing, however.

10.2.5 Voice over Frame Relay

VoFR is receiving growing attention. Most data-oriented frame relay access devices (FRADs) and routers use the FIFO method of handling traffic. In order to achieve the best voice quality, however, voice frames cannot be allowed to accumulate behind a long queue of data frames. Voice FRADs and routers, therefore, employ traffic prioritization schemes to give preference to voice traffic, thereby minimizing delay. During times of network congestion, one of the easiest ways to relieve congestion is to simply discard frames. In such cases, data rather than voice frames will be discarded first, giving voice a better chance of making it through the network.

Some service providers offer prioritization of PVCs within the frame relay network. Prioritization features on both the CPE and the frame relay network can result in better voice application performance. The CPE ensures that higher priority traffic is offered to the network first, while PVC prioritization within the network ensures that higher priority traffic is delivered from node to node first.

VoFR equipment compresses the voice signal from 64 Kbps to at least 32 Kbps. In most cases, compression to 16 Kbps or even 8 Kbps is possible. Some equipment vendors support dynamic compression options. When bandwidth is available, a higher voice quality is achieved using 32 Kbps, but as other calls are placed or other traffic requires bandwidth, the 16- or 8-Kbps compression algorithm is implemented. Most voice FRADs also support fax traffic. A fax can take up as little as 9.6 Kbps of available bandwidth. VoFR usually allows a company to use its existing phones and numbering plan. In most cases, an internal dialing plan can be set up that allows users to dial fewer digits to connect to internal locations.

A persistent myth about VoFR is that voice calls can be carried free on an existing-frame relay network. In fact, VoFR requires special CPE, entails an increase in the port speed, and possibly an increase of the CIR—all of which have a cost associated with them. Furthermore, to communicate with all other corporate locations, there must be a fully meshed network of PVCs, which increases the overall cost. If the carrier offers SVCs—and not all of them do—fewer virtual circuits are needed, but there is still a cost associated with the SVCs that are needed. Despite all this, many companies choose to run VoFR because it saves them money.

Another issue to contend with when considering VoFR is that many carriers refuse to offer it as a separate service and do not address it in their SLAs. Consequently, a company will have to take responsibility for choosing the right CPE and setting it up to handle voice calls, even if that means hiring a consultant. To many carriers, voice is simply another data application running over their frame relay network; if the customer is not satisfied with the QoS for voice, the carrier does not accept responsibility. All this will be discussed prior to subscribing to a frame relay service, so these issues should come as no surprise.

---