CHAPT ER

1 3

Frame Relay
Background
Frame Relay was originally conceived as a protocol for use over ISDN interfaces. Initial proposals to this effect were submitted to the International Telecommunication Union Telecommunication Standardization Sector (ITU-T) (formerly the Consultative Committee for International Telegraph and Telephone [CCITT]) in 1984. Work on Frame Relay was also undertaken in the American National Standards Institute (ANSI)-accredited T1S1 standards committee in the United States. There was a major development in Frame Relays history in 1990 when Cisco Systems, StrataCom, Northern Telecom, and Digital Equipment Corporation formed a consortium to focus Frame Relay technology development and accelerate the introduction of interoperable Frame Relay products. This consortium developed a specication conforming to the basic Frame Relay protocol being discussed in T1S1 and ITU-T, but extended it with features that provide additional capabilities for complex internetworking environments. These Frame Relay extensions are referred to collectively as the local management interface (LMI).

Technology Basics
Frame Relay provides a packet-switching data communications capability that is used across the interface between user devices (for example, routers, bridges, host machines) and network equipment (for example, switching nodes). User devices are often referred to as data terminal equipment (DTE), while network equipment that interfaces to DTE is often referred to as data circuit-terminating equipment (DCE). The network providing the Frame Relay interface can be either a carrier-provided public network or a network of privately owned equipment serving a single enterprise. As an interface to a network, Frame Relay is the same type of protocol as X.25 (see Chapter 12, X.25). However, Frame Relay differs signicantly from X.25 in its functionality and format. In particular, Frame Relay is a more streamlined protocol, facilitating higher performance and greater efciency. As an interface between user and network equipment, Frame Relay provides a means for statistically multiplexing many logical data conversations (referred to as virtual circuits) over a single physical transmission link. This contrasts with systems that use only time-division-multiplexing (TDM) techniques for supporting multiple data streams. Frame Relays statistical multiplexing provides more exible and efcient use of available bandwidth. It can be used without TDM techniques or on top of channels provided by TDM systems. Another important characteristic of Frame Relay is that it exploits the recent advances in wide-area network (WAN) transmission technology. Earlier WAN protocols such as X.25 were developed when analog transmission systems and copper media were predominant. These links are much less reliable than the ber media/digital transmission links available today. Over links such as these, link-layer
Frame Relay 13-1

LMI Extensions

protocols can forego time-consuming error correction algorithms, leaving these to be performed at higher protocol layers. Greater performance and efciency is therefore possible without sacricing data integrity. Frame Relay is designed with this approach in mind. It includes a cyclic redundancy check (CRC) algorithm for detecting corrupted bits (so the data can be discarded), but it does not include any protocol mechanisms for correcting bad data (for example, by retransmitting it at this level of protocol). Another difference between Frame Relay and X.25 is the absence of explicit, per-virtual-circuit ow control in Frame Relay. Now that many upper-layer protocols are effectively executing their own ow control algorithms, the need for this functionality at the link layer has diminished. Frame Relay, therefore, does not include explicit ow control procedures that duplicate those in higher layers. Instead, very simple congestion notication mechanisms are provided to allow a network to inform a user device that the network resources are close to a congested state. This notication can alert higher-layer protocols that ow control may be needed. Current Frame Relay standards address permanent virtual circuits (PVCs) that are administratively congured and managed in a Frame Relay network. Another type, switched virtual circuits (SVCs), has also been proposed. The Integrated Services Digital Network (ISDN) signaling protocol is proposed as the means by which DTE and DCE will communicate to establish, terminate, and manage SVCs dynamically. For more information on ISDN, see Chapter 10, Integrated Services Digital Network. Both T1S1 and ITU-T have work in progress to include SVCs in Frame Relay standards.

LMI Extensions
In addition to the basic Frame Relay protocol functions for transferring data, the consortium Frame Relay specication includes LMI extensions that make supporting large, complex internetworks easier. Some LMI extensions are referred to as common and are expected to be implemented by everyone who adopts the specication. Other LMI functions are referred to as optional. A summary of the LMI extensions follows:

Virtual circuit status messages (common)Provide communication and synchronization between the network and the user device, periodically reporting the existence of new PVCs and the deletion of already existing PVCs, and generally providing information about PVC integrity. Virtual circuit status messages prevent the sending of data into black holes, that is, over PVCs that no longer exist. Multicasting (optional)Allows a sender to transmit a single frame but have it delivered by the network to multiple recipients. Thus, multicasting supports the efcient conveyance of routing protocol messages and address resolution procedures that typically must be sent to many destinations simultaneously. Global addressing (optional)Gives connection identiers global rather than local signicance, allowing them to be used to identify a specic interface to the Frame Relay network. Global addressing makes the Frame Relay network resemble a local-area network (LAN) in terms of addressing; address resolution protocols therefore perform over Frame Relay exactly as they do over a LAN. Simple ow control (optional)Provides for an XON/XOFF ow control mechanism that applies to the entire Frame Relay interface. It is intended for those devices whose higher layers cannot use the congestion notication bits and that need some level of ow control.

13-2 Internetworking Technology Overview

Frame Format

Frame Format
The Frame Relay frame is shown in Figure 13-1. The ags elds delimit the beginning and end of the frame. Following the leading ags eld are two bytes of address information. Ten bits of these two bytes make up the actual circuit ID (called the DLCI, for data link connection identier).

Figure 13-1
Field length, in bytes 1

Frame Relay Frame

The 10-bit DLCI value is the heart of the Frame Relay header. It identies the logical connection that is multiplexed into the physical channel. In the basic (that is, not extended by the LMI) mode of addressing, DLCIs have local signicance; that is, the end devices at two different ends of a connection may use a different DLCI to refer to that same connection. Figure 13-2 provides an example of the use of DLCIs in nonextended Frame Relay addressing.

Figure 13-2
San Jose

Frame Relay Addressing

Router

Switch

Switch

Switch WAN Router Los Angeles DLCI = 12

Switch

Router DLCI = 82 Atlanta

In Figure 13-2, assume two PVCs, one between Atlanta and Los Angeles, and one between San Jose and Pittsburgh. Los Angeles uses DLCI 12 to refer to its PVC with Atlanta, while Atlanta refers to the same PVC as DLCI 82. Similarly, San Jose uses DLCI 12 to refer to its PVC with Pittsburgh. The network uses internal proprietary mechanisms to keep the two locally signicant PVC identiers distinct. At the end of each DLCI byte is an extended address (EA) bit. If this bit is one, the current byte is the last DLCI byte. All implementations currently use a 2-byte DLCI, but the presence of the EA bits means that longer DLCIs may be agreed upon and used in the future. The bit marked C/R following the most signicant DLCI byte is currently not used.
Frame Relay 13-3

S1318a

Flags

Address

Data

FCS

Flags

LMI Message Format

Finally, 3 bits in the 2-byte DLCI provide congestion control. The forward explicit congestion notication (FECN) bit is set by the Frame Relay network in a frame to tell the DTE receiving that frame that congestion was experienced in the path from source to destination. The backward explicit congestion notication (BECN) bit is set by the Frame Relay network in frames traveling in the opposite direction from frames encountering a congested path. The notion behind both of these bits is that the FECN or BECN indication can be promoted to a higher-level protocol that can take ow control action as appropriate. (FECN bits are useful to higher-layer protocols that use receiver-controlled ow control, while BECN bits are signicant to those that depend on emitter-controlled ow control.) The discard eligibility (DE) bit is set by the DTE to tell the Frame Relay network that a frame has lower importance than other frames and should be discarded before other frames if the network becomes short of resources. Thus, it represents a very simple priority mechanism. This bit is usually set only when the network is congested.

LMI Message Format

Figure 13-3
Field length, in bytes

LMI Message Format

Variable

In LMI messages, the basic protocol header is the same as in normal data frames. The actual LMI message begins with four mandatory bytes, followed by a variable number of information elements (IEs). The format and encoding of LMI messages is based on the ANSI T1S1 standard. The rst of the mandatory bytes (unnumbered information indicator) has the same format as the LAPB unnumbered information (UI) frame indicator with the poll/nal bit set to zero. For more information about LAPB, see the section Layer 2 in Chapter 12, X.25. The next byte is referred to as the protocol discriminator, which is set to a value that indicates LMI. The third mandatory byte (call reference) is always lled with zeros. The nal mandatory byte is the message type eld. Two message types have been dened. Status-enquiry messages allow the user device to inquire about network status. Status messages respond to status-enquiry messages. Keepalives (messages sent through a connection to ensure that both sides will continue to regard the connection as active) and PVC status messages are examples of these messages and are the common LMI features that are expected to be a part of every implementation that conforms to the consortium specication. Together, status and status-enquiry messages help verify the integrity of logical and physical links. This information is critical in a routing environment because routing algorithms make decisions based on link integrity. Following the message type eld is some number of IEs. Each IE consists of a single-byte IE identier, an IE length eld, and one or more bytes containing actual data.
13-4 Internetworking Technology Overview

S1320a

Flag

LMI DLCI

Unnumbered Protocol Call Message Information information elements discriminator reference type indicator

FCS

Flag

LMI Message Format

Global Addressing
In addition to the common LMI features, there are several optional LMI extensions that are extremely useful in an internetworking environment. The rst important optional LMI extension is global addressing. As noted earlier, the basic (nonextended) Frame Relay specication only supports values of the DLCI eld that identify PVCs with local signicance. In this case, there are no addresses that identify network interfaces, or nodes attached to these interfaces. Because these addresses do not exist, they cannot be discovered by traditional address resolution and discovery techniques. This means that with normal Frame Relay addressing, static maps must be created to tell routers which DLCIs to use to nd a remote device and its associated internetwork address. The global addressing extension permits node identiers. With this extension, the values inserted in the DLCI eld of a frame are globally signicant addresses of individual end-user devices (for example, routers). This is implemented as shown in Figure 13-4.

Figure 13-4
San Jose

Global Addressing Exchange

Router

DLCI = 12

DLCI = 13

Router

Switch

Switch

Switch

Switch

Switch

WAN
S1321a

Router DLCI = 14 Los Angeles DLCI = 15

Router Atlanta

In Figure 13-4, note that each interface has its own identier. Suppose that Pittsburgh must send a frame to San Jose. The identier for San Jose is 12, so Pittsburgh places the value 12 in the DLCI eld and sends the frame into the Frame Relay network. At the exit point, the DLCI eld contents are changed by the network to 13 to reect the source node of the frame. Each router interface has a distinct value as its node identier, so individual devices can be distinguished. This permits adaptive routing in complex environments. Global addressing provides signicant benets in a large, complex internetwork. The Frame Relay network now appears to the routers on its periphery like any LAN. No changes to higher-layer protocols are needed to take full advantage of their capabilities.

Frame Relay 13-5

Network Implementation

Multicasting
Multicasting is another valuable optional LMI feature. Multicast groups are designated by a series of four reserved DLCI values (1,019 to 1,022). Frames sent by a device using one of these reserved DLCIs are replicated by the network and sent to all exit points in the designated set. The multicasting extension also denes LMI messages that notify user devices of the addition, deletion, and presence of multicast groups. In networks that take advantage of dynamic routing, routing information must be exchanged among many routers. Routing messages can be sent efciently by using frames with a multicast DLCI. This allows messages to be sent to specic groups of routers.

Network Implementation
Frame Relay can be used as an interface to either a publicly available carrier-provided service or to a network of privately owned equipment. A typical means of private network implementation is to equip traditional T1 multiplexers with Frame Relay interfaces for data devices, as well as non-Frame Relay interfaces for other applications such as voice and video-teleconferencing. Figure 13-5 shows this conguration.

Figure 13-5

Hybrid Frame Relay Network

Token Ring

Router Frame Relay interface WAN T1 MUX Non-Frame Relay interface Ethernet

Frame Relay interface

Non-Frame Relay interface

PBX

Token Ring

Router

Video/teleconference Ethernet

13-6 Internetworking Technology Overview

S1322a

T1 MUX

Network Implementation

A public Frame Relay service is deployed by putting Frame Relay switching equipment in the central ofces of a telecommunications carrier. In this case, users can realize economic benets from trafc-sensitive charging rates, and are relieved from the work necessary to administer and maintain the network equipment and service. In either type of network, the lines that connect user devices to the network equipment can operate at a speed selected from a broad range of data rates. Speeds between 56 kbps and 2 Mbps are typical, although Frame Relay can support lower and higher speeds. Implementations capable of operating over 45-Mbps (DS3) links are expected to be available soon. Whether in a public or private network, the support of Frame Relay interfaces to user devices does not necessarily dictate that the Frame Relay protocol is used between the network devices. No standards for interconnecting equipment inside a Frame Relay network currently exist. Thus, traditional circuit-switching, packet-switching, or a hybrid approach combining these technologies can be used.

Frame Relay 13-7

Network Implementation

13-8 Internetworking Technology Overview