Congestion Control with Frame Relay Interface Ports

Introduction

There are two types of congestion notification used to control station transmission rates; Explicit
and Implicit.
• Explicit Congestion notification is done by the attached network sending frames to the Frame Relay Interface station with the Backward Explicit Congestion notification (BECN) bit set in the frame header. This notifies the Frame Relay Interface station that the network is congested for the corresponding DLCI. Frame Relay Interface stations can be configured to respond to Explicit Congestion notification.
• Implicit Congestion notification is the process of an Annex G station detecting lost frames. Frame loss is detected when the LAP-B Annex G station is forced to retransmit a frame. Only Annex G stations can respond to Implicit Congestion notification.
Under normal conditions, neither the Annex G nor the Bypass stations set the
Discard Eligible (DE) bit.

Data Rate

An FRI station normally sends frames at the maximum rate available (line speed). It is possible for the station to exceed its committed rate. Usually, this is a temporary situation and statistically the station sends at or below its committed rate. However, if the network is experiencing congestion, then the implicit or explicit congestion mechanism causes the station to enter a controlled send state and lower its rate of transmission to cooperate with the network in congestion control.
An Annex G station that is in a controlled send state, sends frames that carry voice traffic. This occurs even if transmission causes the send rate to exceed the controlled rate. These frames are not buffered and are sent as quickly as possible. Excess rate voice frames are sent with the DE bit set.

Congestion Control for DTE

For a Frame Relay DTE Interface port, there are five configurable parameters
related to congestion control:
• Committed Information Rate (CIR)
• Committed Burst Size (BC)
• End-to-End Delay
• Congestion Control Mode
• Maximum Information Rate (MIR)

Committed Information Rate (CIR) and Committed Burst Size (BC)

The values to use for the CIR and BC are those to which the Frame Relay port and its DLCIs have subscribed. If this port connects to a Frame Relay carrier, these parameter values are provided by the carrier and should be set accordingly.
•Note
These parameters cannot be tuned; they are set by the provider of the Frame Relay network at subscription time.

End-to-End Delay

The End-to-End delay parameter determines the value of the internal step count parameter used to reduce the transmission rate when congestion is measured by the station. The End-to-End Delay value can be estimated and supplied by the provider of the Frame Relay service. It can also be measured, but this is difficult to do and the estimate is usually sufficient.
These parameters are configured on a per station basis. Excessive frame loss due to congestion indicates the step count used in reducing the transmission rate may be too large. This situation can be improved by adjusting the End-to-End Delay parameter.

Congestion Control Mode
You use the Congestion Control Mode parameter to define how the station handles congestion notification. The FRI station detects the Frame Relay network congested state when it receives a frame from the network with BECN bit set to one (1). As a sender, it constantly monitors this bit in frames received from the Frame Relay network. If the BECN bit is detected as being set, the transmitter reduces its rate of transmitting data bits. Note that the rates are the maximum rates of transmission. Obviously, such rates are achieved only if the transmitter has data constantly queued for transmission on the station. You use the Congestion Control Mode parameter to control the handling of congestion notification for both explicit and implicit congestion notification.

Maximum Information Rate (MIR)

Introduction

In order to control the station outgoing information rate and provide traffic shaping capabilities, a station configuration parameter Maximum Information Rate (MIR) has been created. The MIR parameter is accessible only when Frame Relay station configuration parameter Congestion Control Mode is configured as NORMAL, DISABLE or LIMIT. Valid values for this parameter are between CIR and the local interface access rate. While a network is uncongested, the station maximum average transmission rate is determined by this parameter. Measurement Interval Tc is forced to be in range 50 to 200ms.This reduces burstiness and further reduces a chance for congestion. Large Tc values can cause large gaps between packets, because packets are sent at the beginning of the interval. Smaller Tc values smooth traffic by
spreading one big burst over several time intervals. This reduces the chance for long delays of voice packets caused by previously accumulated data packets in network switches. When the MIR parameter is set to the default value 0, Traffic Shaping is disabled and the station rate and operation are equal to the existing rate and operation. When MIR is enabled station state is Controlled.

•Note
Voice packets and packets having priority PRI_EXP_DROP, will be excepted by the rate control. The packets will not be queued or discarded even when the rate is higher than MIR.

Explicit Congestion Control

Introduction

Both Annex G and Bypass stations permit the use of explicit congestion control. Explicit Congestion Control is the process of reducing a station’s transmission rate when the attached network sends frames to the FRI port, with the BECN bit set. The Congestion Control Mode parameter determines how a station reacts to the BECN.

Normal Congestion Control

This mode of congestion control is obtained by setting the parameter Congestion Control Mode to NORMAL. A station is initially in the uncontrolled state and can transmit data when data is available. This means that the maximum number of characters allowed is only limited by the link speed. Upon receiving the first BECN from the network, the allowed transmission rate is immediately reduced to ensure the CIR is not exceeded, and the station goes into a controlled state. In the controlled state, a step count algorithm calculates two parameters:
Step Count equals (CIR x End-to-End Delay) / max packet size
Delta-T equals Committed Burst Size / CIR

Where max packet size equals a nominal value of 2088 bits, the other values are taken from the stations configured values with the CIR value in bits per second and End-to-End Delay in seconds.

•Note
Step Count cannot be less than 4 or grater than 255

These parameters are used to measure and control congestion and to either reduce the
rate further or increase the rate (re-enter uncontrolled state). Delta-T is the average
time in which a specific number of characters are allowed to be transmitted. While in the controlled state:
• If the number of additional BECNs received (consecutive frames with the BECN bit set) is greater than, or equal to the Step Count, the maximum transmission rate allowed is reduced to 5/8 of CIR. This applies if the allowed rate is between 5/8 CIR and CIR.
• If the number of additional BECNs received (consecutive packets with the BECN bit set) is greater than, or equal to the Step Count, the maximum transmission rate allowed is reduced to 1/2 of CIR. This applies if the allowed rate is between 1/2 CIR and 5/8 CIR.
• If the number of BECNs received (consecutive packets with the BECN bit set) is greater than, or equal to the Step Count, the maximum transmission rate allowed is reduced to 1/4 of CIR. This applies if the allowed rate is between 1/4 CIR and 1/2 CIR.
• If further BECN bits are received, the transmission rate is not set below 1/4 CIR (that is, the lowest transmission rate that can be set).
The Frame Relay network stops sending frames with the BECN bit set when it recovers from its congested state. The FRI station counts the number of consecutive frames received without the BECN bit set. When the number of frames with BECN set to zero exceeds (Step Count)/2, it increases the allowed transmission rate in increments of 1/8 CIR and again counts the number of consecutive frames with BECN set to zero to repeat the increment process. Once the transmission rate reaches CIR, the network leaves the controlled state.

The NORMAL mode is used in most cases when the port is attached to a Frame Relay network provider. It gives a measure of protection from frame loss if the Frame Relay network becomes so congested that it loses frames even if the transmission rate was near CIR bounds. If lost frames requires retransmission, then this is the best mode since retransmission into a congested network causes further congestion.

Disable Congestion Control

This mode of congestion control is obtained by setting the Congestion Control Mode to DISABLE. This disables the FRI station rate reduction congestion management mechanism and you can use this value when frame loss by the network is not an issue. It allows the transmitter to send at its highest rate without regard to possible congestion frame loss. If frame loss is an issue with the application using this DLCI, it usually employs a retransmission scheme to detect and resend lost frames. If this is the case, be aware that disabling congestion control may actually reduce throughput. Retransmissions into an already congested network only adds to the congestion, and congestion likely becomes so severe that overall throughput goes below a level that would be achieved if the transmitter reduced its rate using the congestion notification mechanisms.
•Note
Annex G stations always operate with a LAP-B procedure and retransmit on detecting frame loss. This mode might not be desirable for such stations.

Congested Congestion Control

This mode of congestion control is obtained by setting the Congestion Control Mode to CONG. A station is always in the controlled mode, that is, the maximum transmission rate allowed never exceeds the CIR. In this controlled state, the same rate control algorithm applied in the NORMAL mode is used to further control the transmission rate if BECN bits are received.
This mode allows the transmission rate to be set to a maximum of CIR, even when there is constant data queued for transmission. This mode is useful in situations where the attached network discards frames which are received at a rate greater than CIR.
•Note
An important example of the use of this mode would be a Frame Relay network configured to discard frames that are in excess of CIR.

Limit Congestion Control

This mode of congestion control is obtained by setting the Congestion Control Mode to LIMIT. A station is initially in the uncontrolled state, that is, the maximum transmission rate allowed is limited only by the link speed. Upon receiving the first BECN from the network, the maximum transmission rate allowed is reduced to CIR, and the station goes into the controlled state. The maximum allowed transmission rate is never reduced below CIR, regardless of the number of BECN bits received. Upon receiving [(Step Count)/2] consecutive frames without the BECN bit set, the station goes back into the uncontrolled state.
This mode can be selected when the Frame Relay network is not usually subjected to congestion conditions. Occasional light congestion experienced by the network causes the station to reduce the transmission rate to the CIR value and no lower.

Implicit Congestion Control
Introduction For implicit congestion control on Annex G stations, any time the station needs to
retransmit (frame loss / REJ), it informs the FRI port congestion control mechanism. When congestion control receives this indication, it immediately reduces the maximum allowed transmission rate to 1/4 of CIR and goes into the controlled state. The same rate recovery algorithm used for NORMAL congestion control is used to get out of the controlled state. This consists of receiving [(Step Count)/2] consecutive packets from the network with the BECN bit clear to increase the allowed transmission rate by 1/8 of CIR, and repeating this process until the CIR rate is achieved. A final [(Step Count)/2] count of frames with BECN set to zero moves
the station into the uncontrolled state.
Annex G stations using implicit congestion also use explicit congestion control. In effect, the detection of frame loss is simply an additional method of sending a station into the controlled state, at an allowed rate of 1/4 CIR. Once the station enters a controlled state, the recovery process is the same, that is, counting consecutive frames with BECN set to zero.