Table of Contents *

Virtually everyone who makes data communications their business today has heard of Frame Relay. Most understand what Frame Relay is: an efficient form of packet switching. Frame Relay is one of the newer protocols in the communications world, yet its roots are in well-known and well-proven packet switching techniques. Packet switching is the basis of Frame Relay, and was introduced into business communications some twenty years ago, in the form of Systems Network Architecture (SNA) from IBM, X.25 as defined by the International Standards Organization (ISO), the International Telecommunications Union (ITU), and also the ARPAnet, progenitor of the Internet.

The form of packet switching called Frame Relay emerged from a convergence of several factors:

1. The need for higher speed packet technology driven by LAN internetworking.
2. The need to integrate traffic from legacy and LAN applications over the same links.
3. The availability of a definition for a streamlined packet technique from Integrated Services Digital Network (ISDN).
4. The desire of equipment vendors and communications carriers to deliver a high speed packet service for wide area communications as simply and as rapidly as possible.

Frame Relay is a Datalink Layer protocol that is built on the existing CCITT X.25 and ISDN standards. Frame relay is often thought of as the next generation packet network succeeding X.25. X.25 was developed about 25 years ago and was designed for the use with "noisy" analog lines. To compensate for the error caused by "noisy" data links, X.25 checks and corrects the data traversing every data path. This extensive error checking adversely effects network performance as compared to frame relay. Frame relay realizes its efficiencies and only performs error calculations at the source and destination devices, rather than at each switching node. Error checking at each node is not necessary because today’s digital networks are relatively error-free, as compared to those on analog networks. Since the error verification and correction burden is removed, network response time is greatly improved.

Frame relay is a high-speed packetized data service that consists of physical and logical components. The physical components include Frame Relay Assemblers/Disassemblers (FRAD’s), circuits, and frame relay ports. The logical components consist of Permanent Virtual Circuits (PVC’s). Frame Relay Assemblers/Disassemblers (FRAD’s) are devices, such as routers, that assemble data into frame relay packets and transmit those packets through the local access circuits to which they are connected.

Frame relay’s logical component differentiates it from the traditional high-speed data service that is physical in nature. The logical component of frame relay consists of Permanent Virtual Circuits (PVC’s). PVC’s are the logical connections that allow the various sites, having physical network access, to communicate with one another. Each logical connection defined in the frame relay provider’s network and in the FRAD equipment. The FRAD must be programmed to send the data to the correct PVC. Every PVC provisioned in the frame relay network must also be programmed into the FRAD. FRAD’s are programmed using Datalink Connection Identifiers (DCLI’s) that associate an identification number to each PVC. Each PVC has a Committed Information Rate (CIR) which is the bandwidth associated to the logical connection. Frame relay networks have the ability to transmit data at a rate higher than that of the CIR. This allows a "bandwidth-on-demand" feature that improves the performance of bandwidth intensive applications.

Frame relay has become the data technology of choice for organizations around the world implementing networks at speeds of T1/E1 and below. New generation applications and the growing demand on corporate information systems is creating the need for yet more bandwidth, especially in high traffic areas of the network. Recently, the Frame Relay Forum amended the User-to-Network Interface (UNI) and Network-to-Network Interface (NNI) Implementation Agreements (IAs) to meet user demands for access speeds up to 45 Mbps (DS3). With higher-speed access, users can concentrate more remote locations using fewer access facilities from their service provider.

Frame relay is a datalink layer protocol, which occupies layer 2 of the OSI model (see figure below). To fully understand where frame relay fits in the OSI model, you must first understand the functions of the bottom four layers of the OSI model. Layers 1 through 4 are the primary communication layers.

One of the most popular benefits of frame relay is the vast savings over leased or private lines. Since a frame relay implementation reduces number of dedicated devices, this reduces operations and administrative and equipment costs. Frame relay reduces the complexity of a network. It offers greater bandwidth flexibility, higher reliability and resiliency than that of leased lines. Frame relay is especially attractive to smaller implementation because of the lower cost of ownership and better bandwidth utilization. It is one of the best services on the market today to improve application performance and
network efficiency.

Another of frame relay’s strengths is its flexibility. More direct connectivity between locations can be provisioned for minimal incremental cost. Because of this, frame relay networks can be designed to better match the underlying traffic patterns. Locations and connections can be added easily and more cost effectively with frame relay compared with leased lines. Because changing port connection and virtual circuit bandwidth is software configurable, it is easy to redesign and optimize the network. Most frame relay networks have self-healing or automatic rerouting capabilities between the frame relay switches. Frame relay is based on statistical multiplexing where bandwidth can be shared between active applications and/or connections only. This lowers the cost of ownership because end users don’t have to pay for idle or excess capacity needed to meet peak traffic periods. In a branch office network environment these cost savings can be substantial.

Frame relay circuits are now generally less expensive that equivalent private lines. For example, a coast-to-coast 56Kbps AT&T private line that costs $1800 can be replaced by as equivalent Frame Relay virtual circuit costing less than $1000. Cost factors have triggered a rush to replace private lines with Frame Relay.

Frame relay service is particularly well suited to bursty LAN traffic. Between private-line replacement at larger sites and new installations at smaller sites, frame relay is rapidly becoming the dominant wide-area networking service. Taking advantage of the popularity of frame relay, hardware vendors have introduced a variety of frame relay products, such as FRAD’s and FRAR’s. For network managers considering a move from leased-line networks to frame relay, the big question isn’t whether but how. Frame relay’s main advantages over leased-lines are lower costs and increased bandwidth flexibility. The challenge lies in figuring out what to look for in frame relay access devices.

The bottom line is that users adopted frame relay to save money. However, now that frame relay networks are becoming more popular, other issues must be addressed, such as, scalability, security, and data reliability and accuracy.

In addition to money saving advantages, frame relay now offers a very smooth migration to ATM, another emerging technology. Frame relay is an efficient protocol that has been streamlined to eliminate the link by link flow and error control.

The frame relay structure ordinarily contains the following:

1. Starting Flag (1 octet)
2. Address (2 octets)
3. Information Field (Variable length – up to 4096 octets)
4. Frame Check Sequence (2 octets)
5. Ending Flag ( 1 octet)

The two address octets contain five very unique elements:

1. Data Link Connection Identifier (DLCI) -- This contains the identification of the PVC used by the network to find the right path and destination for the frame.
2. Command/Response Field (C/R) -- Not used.
3. Extended Address (EA) -- Can be used for extended addressing needs.
4. Explicit Congestion Notification Fields (FECN & BECN) -- Used for flow control.
5. Discard Eligibility -- Frames

Frame relay standards define three types of logical circuits:

• Permanent Virtual Circuits (PVC’s)
• Switched Virtual Circuits (SVC’s)
• Multi-cast

The PVC is the primary way in which carriers currently provide service. SVC’s are not commonly available from frame relay service providers at this time, but have several distinct advantages. Finally multi-cast is only provided on a proprietary basis.

A PVC is exactly what is sounds like. This type of connections is permanent and dedicated. PVC’s must be staticly defined at configuration, unless PVC parameters need to be modified. The connection is constantly configured whether there is information to send or not. One limitation imposed on PVC’s is the provisioning of N*(N-1)/2 connections.

In some carrier’s networks, status indicators which reflect the current utilization of the network resources, are updated at every node crossed by each PVC in the network. Depending on the particular values of the status indicators, the algorithm increases, decreases, fast decreases, or does not change the rate of the PVC. However, the indicators will not cause the available bandwidth of the PVC to be decreased below the CIR. The bandwidth of the PVC is adjusted in one of the following ways:

• Rate up: The bandwidth of the PVC is increased as a percentage of the CIR. This occurs when network capacity exists to carry the increased traffic load.
• Rate Down: The bandwidth of the PVC is decreased as a function of the current value. This occurs when there is a high level of traffic on the network.
• Fast Decrease: The bandwidth of the PVC is decreased from its current value by a large percentage. This occurs when there is a very high level of traffic in the network. It will not drop below the CIR.

The bandwidth of the PVC is adjusted by changing the rate at which credits are delivered to the PVC’s credit buffer. Unused credits stored in the credit buffer provide the burst capability. By adjusting the rate at which credits are received, the bandwidth of the PVC is adjusted. The fact that the bandwidth of the PVC’s are increased in proportion to the CIR leads to the fairness property. The unused network bandwidth is shared between the PVC’s in proportion to their particular CIR’s.

Switched Virtual Circuits (SVC’s) enable frame relay users to dynamically expand their current PVC networks and establish logical network connections on an as-needed basis to end points on the same network or through gateways to end points on other networks. Benefits to users include improving network performance, further reducing network costs and simplifying network management.

Several applications are driving the need for SVC’s such as intranets, desktop video conferencing, remote access and voice applications. These types of applications typically do not require the dedicated bandwidth capabilities of PVC’s because connections are only occasionally needed. Using SVC’s, users can access information or communicate directly over dynamic connections that do not require pre-configuration. In fact, SVC’s enable users to cost-effectively expand their existing PVC-based frame relay networks.

SVC’s and PVC’s can co-exist in the same network. PVC’s are better suited to end points that frequently communicate. SVC’s, on the other hand, are ideal for occasional-use applications or end points that need to periodically communicate. Additionally, standards-based SVC’s can easily create connections across the NNI (network-to-network interface) to end points on other frame relay networks. Together, PVC’s and SVC’s provide frame relay solutions that are scaleable, reliable and cost-effective.

Switched Virtual Circuits (SVC’s) and Permanent Virtual Circuits (PVC’s) both enable logical connections. PVC networks are typically configured using a star topology and require pre-configuration of the addresses, bandwidth parameters and traffic priorities. SVC’s, on the other hand, establish temporary point-to-point connections to any addressable node connected to the frame relay network. This provides greater flexibility, as well as reduced set-up and management costs.

SVC’s allow users to establish reliable connections to any destination at any time for any duration desired. Service providers are not required to pre-configure SVC’s, unlike PVC’s, because the user provisions a connection simply by placing a call. SVC’s can also be used as a backup in the event of a transmission system failure because they are available on demand, similar to dial modems and ISDN.

Frame relay has a feature called the Committed Information Rate (CIR) that is designed to accommodate bursty traffic in an efficient manner. The CIR is the rate that is "guaranteed" to a particular subscriber on a particular DLCI. The physical port to the subscriber is usually set at a rate higher than the CIR. If the data source tries to send data at a rate higher than the CIR, the network attempts to send the excess burst on a best effort basis. There can be multiple DLCI’s in one physical link, each one with individual CIR’s. Because multiple PVC’s can occupy the same physical link, any one of the PVC’s can use the full bandwidth of the access link, in excess of the actual CIR of the virtual circuit. For example, each of four PVC’s can have a CIR of 9600 bps, any one of them could use the full 56kbs of the access facility if the remaining PVC’s are idle. However, the risk involved is that bursty frames offered to the network in excess of the CIR are marked as Discard Eligible (DE). As the network becomes congested, these DE frames may be discarded to ensure that the network does not become overloaded.

Many inexpensive frame relay services are based on a CIR of zero. This means that every frame is a DE frame, and the network will discard any or all frames if congestion occurs!

There are two basic methods for dealing with errors or congestion: Automatic Repeat Request (ARQ) and discard. The ARQ error correction scheme acknowledges all frames as good or bad. It keeps all good frames and retransmits the bad ones. With the widespread deployment of fiber, the datalink layer protocols have made a practice of discarding bad frames and allowing the transport layer to provide the recovery mechanism. This allows the datalink layer to be simple and operate with low delay.

Since frame relay networks offer only minimal safeguards against congestion, congestion control is especially important. With frame relay service, customers pay for a Committed Information Rate (CIR), which guarantees the theoretical minimum amount of bandwidth of a given installation. A large advantage to frame relay is that users can utilize bandwidth above the CIR as it is need, provided the extra bandwidth is available. However, is the network suddenly become congested, frame relay switches may respond by discarding frames. A common practice used to prevent frame loss it to over provision the network needs. Although there is no guarantee that frame relay networks will always have ample capacity, as more users access the frame relay services, the odds of congestion increase. When network traffic increases, frame relay switches mark frames with FECN or BECN bits to notify access devices that the network is becoming congested. These flags are indicators to other network devices to react accordingly in attempts to reduce congestion.

Frame relay headers also contain a crude form of prioritization, which incorporates the use of a Discard Eligibility (DE) bit. Frames that have their DE bits set are eligible to be dropped if congestion occurs. However, setting the DE bits in some frames doesn’t ensure that other will make it through a congested network, because an immediate switch in the network can set DE bits on all frames above the CIR that pass through it if congestion becomes extreme. The network will do its best to deliver all packets, but will discard and DE packets first if there is congestion.

Network traffic continues to increase as companies expand, adopt new applications and develop transaction-based relationships with customers and business partners. As traffic among these sites continues to grow, the need for SVCs becomes more apparent. SVCs deliver the necessary bandwidth when needed, while allowing users to manage the additional costs of those connections

Congestion control is handled in two ways: congestion avoidance and congestion recovery. Congestion avoidance is made up of one Backward Explicit Congestion Notification (BECN) bit and one Forward Explicit Congestion Notification (FECN) bit. The BECN bit notifies the transmitting node of potential congestion when there is a build-up of queue lengths, informing the user that the flow of frames should be restricted. The FECN bit notifies the receiving node of potential future delays, informing the user to invoke a higher layer protocol that will alert the transmitting node to restrict the flow of frames. Congestion avoidance is particularly important to high-speed frame relay implementations because it helps reduce the amount of in-flight data loss, resulting in higher throughput.

If congestion avoidance is ineffective, the switching node may discard frames as a result of buffer overflow. In the frame relay packet, a Discard Eligibility (DE) bit indicates the most eligible frames to be discarded. When frames are discarded, the higher-layer protocols are responsible for detecting the lost frames and retransmitting them.

The future of frame relay lies in its interoperability with the next generation of high speed packetized data service – Asynchronous Transfer Mode (ATM). The organizations, which organize standards, have ratified a new standard that establishes interoperability issues between frame relay and ATM networks.

Most customers have deployed frame relay in a star topology, similar to their previous leased-line networks. In this configuration, one location functions as the central point of traffic for one to many remote locations. The central point of the star typically requires the most bandwidth because more traffic passes through it than any one of the remote locations.

Initially, customers ordered 64 kbps access (DS0) for remote office locations and 1.55 Mbps (DS1) for the headquarters site. When traffic or number of locations increased and signs of congestion became apparent at the headquarters site, customers were relegated to ordering another DS1 facility from their service providers which doubled access cost and required additional hardware or Customer Premises Equipment (CPE). With the advent of higher speed frame relay, customers can now gain higher-speed access from their service providers without having to make modifications to the CPE or protocols.

Some of the major Internet service providers are deploying high-speed frame relay backbones to guarantee customers' network availability. Increasingly, Internet users are demanding 56 kbps, T1 and even T3 connectivity. To keep pace with these bandwidth requirements, Internet service providers are choosing frame relay to provide high performance, cost-effective solutions to their customers.

High-speed frame relay provides a viable alternative to end users who are not ready to commit to ATM services. With the ability to reach to DS3 speeds, the investment in frame relay equipment and services can be maximized for years to come.

Like the original UNI and NNI specifications, high speed frame relay assumes that intelligent end devices will guarantee the integrity of transmission and that the physical/electrical connections have low error rates. Therefore, error recovery is not required in the intermediate network nodes. Connection control signaling is carried on a separate logical connection, requiring only one connection control channel at the UNI. The result is higher throughput and shorter delays.

Frame relay to ATM Interworking provides a means to seamlessly integrate frame relay and ATM networks. Several Implementation Agreements (IAs) have been endorsed by the ATM Forum and the Frame Relay Forum that make combining frame relay and ATM networks possible. Two IAs have been developed specifically for current frame relay users: Network Interworking (FRF.5) and Service Interworking (FRF.8). Both solutions protect current investments in frame relay while providing a migration path to ATM.

Frame relay is becoming the transmission technology of choice for WAN users around the world because it is economical and highly efficient. Optimized for data traffic, frame relay's explosive growth has been driven by LAN internetworking, SNA migration, remote access and Internet connectivity. Multimedia and other high-bandwidth applications are better suited to ATM where a higher level of service is required. With frame relay to ATM interworking, users can deploy both frame relay and ATM as necessary to suit their particular enterprise networking needs.

Network Interworking allows connection of two frame relay end nodes such as Frame Relay Access Devices (FRAD’s) or routers, which are attached to a frame relay network, over an ATM backbone. The FRAD’s have no knowledge of the ATM backbone because the network equipment, particularly ATM WAN switches, provide the interworking function. Multiple frame relay networks can be supported by an ATM backbone, providing users with a scalable, high-speed option that does not require changes to the customer premises equipment.

Frame relay has enjoyed and is expected to continue to enjoy tremendous success in the United States as shown in this graph. Frame relay is now reaping the benefits of its early success and proven performance with many users as they begin to migrate their private TDM-based networks, mission-critical applications, and multi-protocol applications onto frame relay. Additionally, growing end user networks, demand for higher speeds, rapidly changing end user needs, IT resource rightsizing, and controlled/limited budget have created even more opportunities for frame relay to proliferate.

The one technology to rival frame relay’s extraordinary growth in the networking world is the commercial Internet (defined here as Internet utilized for business purposes; i.e. access, intranet, extranet, etc.). Research companies vary widely in their estimates today, but typically place 1997 revenues for commercial Internet services between $1 and $2 billion with 1998 projections as high as $3 billion.

Frame relay is a technology that can create a robust wide area networking fabric that integrates information systems together to form an enterprise network. It is an affordable and capable service for supporting today’s bandwidth intensive applications. Because logical connections are defined in software, it is easy to manage moves, changes, additions, and deletions of logical connections.

In addition to cost effectiveness, another advantage of frame relay is the ability to handle traffic bursts efficiently. This is very important since today’s primary source of traffic, namely LAN’s create bursty traffic patterns.

Miller, M.A. (1991). Internetworking: A Guide to Network Communications LAN to LAN; LAN to WAN. New York, NY: M&T Books.

Thyfault, M.E. (1995). Baton Passes to Frame Relay: New standard will make it simpler for carriers to connect frame relay networks to ATM. Information Week, 96.

http://scis.nova.edu/~raciti/frame_1.html

http://www.starburstcom.com/frelay.htm

http://www.techguide.com

http://www.myhomee.org/pg/frame.htm

http://www.acclaiminc.com/PI_wpap.htm

http://www.kmj.com/netlink/frad.html

http://www.dcbnet.com/notes/framerly.html

http://www.sangoma.com/fr.html