Frame Relay uses the public switched data network (PSDN) and is thus a packet-switching technology that is in common use for WAN connectivity. Frame Relay differs significantly from the circuit-switched point-to-point technologies, such as the North American T1. The most significant difference is that Frame Relay networks consist of many nodes sharing the same physical network.
The Frame Relay node at the customer location talks to a network switch. This connection is a private line between the customer and the Frame Relay network. The remote location mirrors this configuration.
Circuit-Switched Versus Packet-Switched
Circuit-switched technology provides dedicated bandwidth between endpoints. This makes it significantly easier to provide quality of service (QoS) in terms of voice quality and delay. This same attribute, dedicated bandwidth between endpoints, means that use of network resources is poor.
Packet-switched networks are more economic in resource use. Because there is no dedicated channel that might not be carrying any traffic, the full bandwidth of the circuit is available to carry the packetized data. Unfortunately for real-time communications, the variables of congestion and delay can seriously affect the quality of service of voice and video traffic.
Advantages of Frame Relay
There are some immediately obvious advantages to considering Frame Relay, for both the customer and the ISP.
First and foremost, it is generally less expensive than traditional point-to-point circuits. There are several reasons for this. Consider the following:
You are purchasing a shorter dedicated circuit. The private line between the customer premises equipment (CPE) and the Frame Relay network switches at the local telco.
The ISP can be much farther away. Because Frame Relay uses packet-switching technology and not dedicated circuit-switching technology, Frame Relay is not billed per mile. You are now free to pick the ISP with the best service and price, without being limited to local only choices.
Usually, the ISP already has an existing CPE. If the ISP already offers Frame Relay, it does not necessarily have to purchase new equipment as it would if supporting a dedicated point-to-point circuit.
The ISP can aggregate many customers onto a single physical circuit. Generally speaking, 40 56Kbps Frame Relay connections can be handled over a single DS1 rate circuit.
The max burst rate permits data to transmit as fast as the circuit's capacity. This works out quite well for short bursts of traffic.
Frame Relay is also widely deployed, which is a major concern for newer technologies such as xDSL and cable.
All these advantages don't come free, however. There are disadvantages to Frame Relay, but they might not affect your requirements for a circuit. The major disadvantage of Frame Relay is caused by the same design that provides the advantages. Because Frame Relay is a packet-switched technology, latency is introduced in the processing of each packet at every Frame Relay switch. This should be negligible for short distances, but it adds up over long distances. A rule of thumb is that for every 1,000 circuit miles, expect 20ms of latency.
Frame Relay is available in any combination of fractional DS1 rates. Common Frame Relay capacities include 128, 256, 384, 512, 768, 1024, and up to 1544Kbps for North American T1, and 2048Kbps for European E1.
Components of Frame Relay
There are five major components of Frame Relay: the data link connection identifier (DLCI), the permanent virtual circuit (PVC), the switched virtual circuit (SVC), the Committed Information Rate (CIR), and the Local Management Interface (LMI).
Data Link Connection Identifier (DLCI)
In every packet-switched network, including Frame Relay, each packet must know its source and destination addresses. This is similar to the Layer 2 data link Media Access Control (MAC) address used in modern LAN technologies. In Frame Relay, the addressing information is contained in the packet header as the DLCI.
The data link connection identifier, or DLCI, is attached to data frames on the Frame Relay network. This information is used to route the frames within the telecom network.
Another use of the DLCI is to differentiate between virtual connections on the same physical port. Packets for several different virtual ports are statistically multiplexed onto the physical port to be transmitted. The aggregated logical connections can coexist on a single physical port because it's the DLCI that logically binds the data to the connection.
Permanent Virtual Circuit (PVC) and Switched Virtual Circuit (SVC)
If you've been paying attention thus far, you know that in a Frame Relay architecture, you have a physical dedicated circuit between your CPE and the telecom's network switch. The same configuration exists at your ISP's location. Because between those two network switches the data is packet-switched, there is no guarantee that the path the data travels is consistent. This is why Frame Relay networks are generally represented by a cloud in diagrams. In reality, the data generally travels the same path; it would travel different routes only if there were significant amounts of congestion or hardware failure.
The path between the telecom switches is commonly referred to as a virtual circuit. This is an apt description because there is no actual dedicated circuit, and your data can travel over many physical paths and switches. Because variety is the spice of life, there is, of course, a choice of virtual circuits.
The switched virtual circuit (SVC) is constructed when there is a need to pass data. After some defined threshold has been reached (such as no traffic for the past 30 seconds), the virtual circuit is torn down. This acts in many respects like the normal analog phone system.
The permanent virtual circuit (PVC), shown in Figure 4, is preconfigured to connect to a specific DLCI. The circuit is always available and always connected. Most ISPs deploy Frame Relay using PVCs.
Figure 4 In a Frame Relay PVC mesh, each router has at least two connections set up in a ring. It's fast, but the problem is that you're paying for each of these lines.
Using PVCs, the ISPs can now use the DLCI to define logical connections to virtual ports in the routers over a single physical circuit.
An extremely useful application of PVCs can be demonstrated when you want to connect multiple sites together directly but cost-effectively. Frame Relay does not have to be deployed in strictly a point-to-multipoint configuration. There are several applications of deploying a mesh of PVCs between remote locations, providing multiple virtual paths and direct routes between remote locations without having to first travel through your ISP's router. You have to pay for each PVC defined through the Frame Relay network, but the performance trade-off might be worth the additional expense. Figure 5 shows an example of a mesh of PVCs.
Figure 5 In a switched virtual circuit, numerous connections exist within the mesh; they are created and destroyed as necessary.
Committed Information Rate (CIR)
The second major difference between Frame Relay and dedicated point-to-point circuits is how the circuit capacity is described. Dedicated point-to-point circuits are just that: dedicated. If you send no data, the channel time slots are still allocated to your circuit.
Remember that Frame Relay is a shared network using packet-switched technology. The intent is to use the full capacity of the circuits, whether or not you're transmitting any data. Therefore, in a shared environment, there needs to be some provision to make sure that you get what you're paying for.
The Committed Information Rate (CIR) has been described as the worst-case throughput that the Frame Relay network provider attempts to guarantee.
The physical port speed is the absolute maximum data rate accepted by the Frame Relay network provider. The port speed is generally twice that of the CIR.
Frame Relay uses the LAPF variant of HDLC as the transmitting protocol used to physically put frames on the Frame Relay network. HDLC is a synchronous protocol, meaning that it is synchronized to a clock source. LAPF and HDLC are defined at Layer 2 of the OSI model. When data is transmitted, the whole frame is transmitted at the clock speed, thus bursting over CIR because there is no way to slow down the data.
The difference between the guaranteed CIR and the port speed is available on a best-effort-only basis.
Local Management Interface (LMI)
The Local Management Interface (LMI) is a keep-alive mechanism to periodically check to make sure that the interface is still active. Additionally, it is used to give the end user circuit status information such as whether the link is congested.
There are three versions of LMI:
LMIThe Frame Relay Forum Implementation Agreement (IA) FRF.1, which has been superseded by FRF.1.1
Annex DAs defined in ANSI T1.617
Annex AITU 933, as referenced in FRF.1.1
Unfortunately, LMI is the generic term used to describe all protocols rather than specifically IA FRF.1. As with practically all standards, the three versions of LMI are not interoperable. LMI is fairly ubiquitous. Most vendors support Annex D, although very few vendors support Annex A.
Congestion and Delay
Delay and congestion are a direct result of the shared nature of packet network resources. Each piece of the packet-switched data network introduces delays because of the processing that occurs at each node to determine the next hop for the packet.
Delay is synonymous with latency, which is the time required for transmission from origin to destination. Latency is usually measured in milliseconds. The threshold for delay interval intolerance for humans is roughly 250ms.
Delay is not a problem in circuit-switched networks because each call uses a reserved circuit. Congestion exists in circuit-switched networks only when all available timeslots are used. You might be familiar with the phrase "All circuits are busy." In a circuit-switched network, this is easily addressable by adding network capacity in appropriate locations.
Shared packet-switched networks are much more dynamic and bursty. Frame Relay networks aggregate multiple PVCs onto a single physical circuit. When there is more aggregated data in the network than there is bandwidth, you have congestion. This is similar to merging five lanes of traffic into two lanes over a bridge. This normally doesn't cause a problem for 20 hours a day, but during rush hour it seems to have an effect on people's blood pressure.
The Frame Relay Forum has defined several mechanisms to handle the congestion problem. Several bits are defined in the Frame Relay header that allow network switches to notify other switches or Frame Relay end nodes of impending congestion (see Figure 6).
Figure 6 Part of the Frame Relay header deals with congestion.
The Forward Explicit Congestion Notification (FECN) bit is added to a congested frame. The receiving Frame Relay network router detects that congestion is occurring when it receives the FECN from the network switch.
The Backward Explicit Congestion Notification (BECN) bit is added to a frame returning to the router that is causing congestion. The transmitting Frame Relay network router detects the congested states when it receives the BECN from the network switch. It then reduces the transmitting rate to the maximum or below CIR until the congestion in the network is reduced. Normal transmission is then resumed.
The Discard Eligibility bit is a self-elective way of saying, "If you have to, you can discard this frame." One method of electing what can be discarded is via prioritization. Prioritization allows the end node to define which traffic is more important and must go through. An example of this might be prioritizing interactive traffic such as Telnet or SSH over noninteractive traffic such as FTP or HTTP.