Read lesson 9b. Then complete hw 9b.

Frame Relay
Frame Relay is the highlight of this lesson.  It is a WAN protocol that runs most commonly on top of ISDN services, such as a T1 connection.  Frame Relay was established as part of the ISDN standardization effort.  However, ISDN is not required to use Frame Relay.  Frame Relay depends on several features of modern networks to make improvements over X.25. 

X.25 had many advantages over older protocols, but there are some disadvantages to using X.25 as well. 

• X.25 sends call control packets on the same link as the data, which increases overhead.
• Flow control is implemented at both layer 3 and layer 2 of the X.25 protocol suite.
• Acknowledgments are also send at both layer 3 and layer 2.

X.25 is extremely reliable and this reliability is needed when networks have high error rates. 

• A reliable protocol is one that requests retransmission of erred data.
• An unreliable protocol is one that does not request retransmission of erred data.

X.25 establishes a virtual circuit through the network.  Frame Relay does this, too.  A connection oriented service uses a call setup procedure to establish this virtual circuit.  So, we can say the main difference between X.25 and Frame Relay as follows:

• X.25 is a reliable connection-oriented service
• Frame Relay is an unreliable connection-oriented service

Current digital networks operate with very low error rates.  All the overhead in an X.25 network is unnecessary.  High quality equipment and high quality lines make the overhead a waste of capacity.  Do we really need to give away capacity to make sure we catch one error per day?  Wouldn't it be better just to let a higher layer take care of the error and request a retransmission, and then reclaim, for example, 10% of our capacity?  Of course it would.

Frame Relay was developed to take advantage of very low error networks.  Here are some highlights of Frame Relay:

• Operates at Layer 2 only
• Uses a separate signaling channel
• Does not keep tables at each intermediate node so there are no control packets in the middle of the network
• There is  no end-to-end flow control and no end-to-end error control
• There are no end-to-end acknowledgments
• less overhead
• lower delay

Wow, we just cut out a whole bunch of overhead.  A Frame Relay frame header should be pretty small.  But wait a minute.  Won't there be some problems if we don't have any flow control or error control?

Potential problems in a Frame Relay Network

• Have to rely on a higher layer protocol for error checking
• Have to rely on higher layers to implement flow control
• End nodes have to use more processor power to implement these functions

Frame Relay communications can be divided into two realms.  The user plane and the control plane.  Communications from end to end for information transfer by the user define the user plane.  Communications between the end user and the Frame Relay network for controlling the network define the control plane. 

Frame Relay Control Plane
The purpose of the control plane is to setup, maintain and tear down the users connection to the network.  Control messages are only significant for the connection between the user and the Frame Relay network.  It is even possible that the user is running the Frame Relay over an ISDN connection.  This could be a BRI connection as described earlier or a PRI connection (T1 connection, 1.544 Mbps).

Running Frame Relay over a BRI ISDN link permits all the control information to be passed on the D channel.  No control information takes up capacity on the informational channels, the B channels.  In this case, the user could get the full 128,000 bits per second for information transfer.

Control messages follow a particular protocol for exchange on the D channel called LAP-D.  LAP-D is an HDLC derivative.  LAP-D highlights follow:

• LAP is short for Link Access Protocol and D stands for D channel
• It is a hybrid of HDLC
• Protocol developed just for encapsulating control information exchanged with the network, such as setup and release messages
• Uses flow and error control, but only for the control messages sent on the D channel

On a T1 line using DS1 framing, 23 channels (B channels) are used for user information transfer and the remaining channel (D channel) is used as a control channel.  In this case, the D channel supports 64 kbps as do the B channels.  The D channel will exchange messages using LAP-D. 

Frame Relay User Plane
Information frames transported across a Frame Relay network use yet another derivative of HDLC, called Link Access Protocol for Frame-mode Bearer Services, or LAP-F.  We'll just call it LAP-F.  This protocol has the the following features:

• Operates at Layer 2 of the OSI model
• Works in ABM mode only (peer  to peer)
• Always uses 7 bit sequence numbers
• Always uses a 16 bit frame check sequence
• NO error control or flow control
• Data Link Connection Identifier, a number used to route the frames through the network, sets up a virtual circuit.
• Frames are, therefore, received in order at the opposite end
• There is a small probability of frame loss

Because there is no error or flow control, the header for this HDLC variant is small.  Take a look at the Frame Relay frame.

Compare this with an HDLC frame.  Here, there is no control field.  The control field was where we had sequence numbers that provided flow control.  The control field also allowed us to send error control messages as well.  While LAP-F checks for errors with a 16-bit CRC, it will not notify the sender if the frame has an error.  Erred frames are discarded.

The address field is used to identify the virtual circuit.  It contains the DLCI and a few other fields.  Take a look at how the address field is broken down for a 2 octet address field, which is the default size.

The address field serves many purposes.  The DLCI identifies the virtual circuit.  Once the packet gets to the Frame Relay network, a frame handler will direct it through the frame relay network using the DLCI.  In a two octet header there are 10 DLCI bits.  That means:

2¹⁰ unique values which means 1024 virtual circuits

For a three octet header (Figure 10.7, pg. 313) there are 17 bits for the DLCI.  For a four octet header there are 24 bits for the DLCI.  These numbers would give us:

2¹⁷ unique values or 131,072 virtual circuits
2²⁴  unique values or 16,777,216 virtual circuits

A DLCI number is determined for use only between the end station and the frame handler at the edge of the Frame Relay network.  The DLCI value is probably different on the other side of the network.  The Frame Relay network holds responsibility to deliver the packets to the other end in order, but how it does this is not supposed to concern the end user.  A variety of protocols might be used to move the frames from the entry to the Frame Relay cloud to the exit from the cloud.

Congestion Management in Frame Relay Networks
Because there is no control field, the standard sliding window flow control cannot be implemented in the network  Without any method of control, congestion could prevent operation of the network. Some method of control is needed.

The network can work together with the end stations during operation to control congestion.  Frame Relay nodes cannot implement direct flow control.  However, three bits are provided in each Frame Relay frame to implement notification and frame discard.  The DE bit is used to mark frames for discard, the FECN bit is used to mark frames that have encountered congestion, and the BECN bit is used to mark frames that pass through routers with congestion in the "backward" direction.

Congestion may be controlled through cooperation:

• Intermediate nodes monitor congestion but, other than throwing frames out which is a drastic type of flow control, can only notify the end stations.
• End stations don't directly see congestion in the network but are responsible to implement flow control to relieve this congestion.
• End stations and intermediate nodes must work together to control congestion in the network.

Another method of control that the Frame Relay provider has is to control the amount of data coming into his network through contractual agreements with the end users.  Frame Relay providers will contract with the end user to provide a Committed Information Rate, or CIR.  In theory, as long as you transmit at or below the CIR, your data travels through the network with a high probability of arriving at the destination.  The CIR is sometimes called the committed burst size, or B_c.

You may exceed the CIR for brief periods of time, but the frame handler will mark your frames as eligible for discard by setting the DE bit to 1.  If this frame encounters congestion it will be thrown away, leaving room for frames sent below the users CIR.  If you continue to exceed the CIR, you will pass the maximum transmission rate, or B_e for excess burst size, for a given period of time.  All frames sent after this are discarded immediately.  You will not be able to send frames again until the (very short) time period expires.

Take care to review a Frame Relay contract carefully.  The
CIR and other values are important.  You may be able to
negotiate these values with the provider.

How is congestion handled when it occurs? 

• At the first signs of congestion, an intermediate node will set the FECN or BECN bits to 1.  This notifies the receiver that congestion is imminent in the network and strong measures need to be taken to limit transmissions.  When the frame comes in with either the FECN bit or the BECN bit set, the receiver will act as follows:

FECN bit set to 1:  receiver should use flow control (higher level) to force sender to slow down.  i.e. withhold ACK messages that open the window or send an receiver not ready (RNR) message to stop the transmitter.

BECN bit set to 1:  receiver should reduce its own transmission rate

• As congestion grows, frames with the DE bit set to 1 are thrown away.  Hey, you were warned this could happen when you signed the contract, right?
• Extreme congestion will force the an intermediate node to drop frames because there is no buffer space to hold them when they come in.  End stations should try to infer congestion by noticing frame loss and then take action. 
• Human operator gets frustrated and ends network session, relieving network of some of its congestion.

The following graphic shows how a frame, with all three congestion notification bits set to zero at the start, is changed to notify the receiver of network congestion.

Review of congestion control bits:

DE bit:  The frame handler at the edge of the network keeps track of your incoming data rate.  Frames that exceed the CIR are marked as discard eligible by setting the DE bit to one.  If a congested intermediate node gets this frame it may, optionally, discard the frame.  No notification is sent to the sender or receiver when a frame is discarded.

FECN bit:  Intermediate nodes set this bit to one if there is congestion moving in the same direction as the frame.  With this bit set, this frame tells the receiver, "your higher level protocol should implement flow control to slow down the incoming frames."

BECN bit: Intermediate nodes set this bit to one if there is congestion in the opposite direction of the packet.  With this bit set, this frame tells the receiver, "frames you send are contributing to congestion, so slow down."

Used with permission of the Author; Copyright (C) Kevin A. Shaffer 1998 - 2022, all rights reserved.