Control congestion on the Frame Relay Network

1. Introduction

Control the occurrence of congestion on the network when overload occurs. There are two possible overload in the network:

1. Call in the new block, and

2. Adjust the situation with the network (create new resources or by reducing the command in the network or by reducing the additional service).

A Frame Relay network to bring the benefits of data "bursty" traffic, as the decision to block a new call is only executed if a combination of capacity on average (not maximum) on the call flow akan dilebihkan. Solution of congestion in the Frame Relay network is trying mengadabtasikan number of entries from frame to frame-in the current congestion. For the "flow control" is not available on the layer-2 user-network interface (in the flow control occurs at the Frame Relay end-to-end), this can not be used to control the "congestion" in some cases, such as packet-switch network. Or in case of congestion, each user must detect congestion is "implicit" (with the use of some services), or when the network detects a congestion situation, the "explicit" must be notified to the user. Users must take action to reduce the number of frame-frame that is inserted in the network.
Congestion occurs when the source of network overload, the source will be the individual transmission link, the buffer is full-on intermediate nodes or the destination system or process in one of these systems. Congestion may also occur because of interference.

That congestion control can not be achieved by adding the resources in the network formation in the buffer capacity or add more high-speed link. Both can not be controlled with configuration balance; the incident can be predicted traffic, congestion can still occur. Congestion may be "inherent" occurs in some packet network and Frame Relay network without exception. Therefore it is important to have a "strategy" to control congestion Frame Relay network.
If the burden of increased traffic, congestion will become increasingly serious (severe) and the burden of the processor system will be more serious, and can lead to system failure. To prevent the occurrence of the failure of this system, a necessary control congestion that can reduce the burden on the system.
2. Indicators and how to overcome congestion
Kurva can be divided into three (3) the level of network congestion. These areas can be considered as indicators of congestion which may be the approach to controlling congestion in each region. Below shows the area does not have congestion. Congestion "moderate" occurred between the "knee" and "cliff", resources can exceed and in accordance with the Qualiy of Service (QOS) (throughput, frame delay or loss) can not be maintained.
By setting the performance of a network congestion scheme can be implemented to overcome the congestion which will take the form of a kurva with two (2) the first pass and prevents the cliff. The plan has a lot of congestion on the object number of entries that diminimumkan Discard the frame, maintaining the QOS, prevent a single-user from the use of monopoly and limit the occurrence of network congestion on the network or on another network. A prevention plan that will be both easy to be implemented and will result in minimal additional network traffic.

3. Approach to control congestion in the Frame Relay network
There are three (3) main approach to control congestion in the Frame Relay network:

1. User to detect the implicit local congestion on the network,

2. Users can mengindikasi network from which the frame-frame may be discarded when congestion occurs, and

3. Notice given to the user at the time of the network itself to detect the congestion.

CCITT Recommendation I.370 provide about congestion management for the ISDN frame relaying bearer service, ie to complete the action associated with the prevention of congestion on the network response and end-user. The recommendations also specify that even if the congestion notification is not appropriate command, this is very much needed.
Users can set the implicit congestion based on a number of parameters as an example the number of samanthk received, the frame rate loss occurred or the throughput of the call. Control is usually used with varying window sizes. For example in the system basic input-which is described later, the difference window to increase / decrease "algorithm" that is used depending on the area where congestion is anticipated.

In plan (scheme) Bandwidth Management (BMW), end-user network with menegosiasi during the call for the establishment of a guarantee throughput (Committed Information Rate = CIR). During phase pentransferan data, monitor the network from the frame rate which is sent by end-users. If the arrival rate corresponding excessive throughput approved, sent some of the frame above the level of the label (using the indicator DE = Discard Eligibility) according to the excess frames. Frame-frame that is sent under the committed level labeled frames. At the time of network congestion occurs, the excess frames are removed. When there is no congestion, end-user may increase the rate of the above negotiations approved. BMW call for the application of the admittance and congestion control used in the ATM network.

Explicitly the basic scheme in the network knows the queue buffer occupancy and the level of which may link the individual overload (overload). Binary Feedback explicit information (EBF) can be used directly on the condition of congestion, and in recommendations I.370 describes the congestion. To control the transfer-destination (a place for the rate-control, which usually works on the transport-layer) to meneruskanpemberitahuan congestion explicitly. In keadaaan source-transfer control (for which the information composing window well) to send the information directly in the backward explicitly notify congestion. Notice that may mengidikasikan location in the frame menormalkan travel directly on the remaining overhead (above) with the technical approach.

Another form of the explicitly considered for the call Stop Duration. In the scheme, the network continuously monitors the return of frame-frame in the buffer. If the placement of the frame can hit the other in the buffer-occupancy exceeds some threshold, the transmission will "stop message" by the network to the sender end-user to answer and stop all traffic on the layer-3. During the parameters taken in determining "stop message" after a period which may be to send more traffic.

4. Prediction techniques to overcome congestion

Techniques to avoid congestion in implicit method is used with the basic input-Kalman filter method, using Kalman Congestion Avoidance Scheme (CAS). Observed throughput at particular virtual circuit can be used to conclude the description of the status of congestion in the network. Delivery user can be used as a preventive action to avoid congestion before actually going, or alternative / other possibilities occur the amount of flow of traffic improvements that are expected to anticipate the load on the network.
This scheme requires 4 (four) steps:

1. Throughput observations.
2. Make predictions congestion level.
3. Decision made as to increase or reduce the burden of traffic.
4. Window added or reduced in line with the firm.
Repeated rotation in some preset interval observation.
5. Conclusion

The best solution for congestion control in the Frame Relay network is one of the basic congestion better than the simple pendeteksian and recovery (recorvery) and this is stipulated in CCITT recommendations.
Basic input techniques we often use the approach for handling congestion, and will work when the network does not provide notification of congestion on the user. This does not require more work with the network and the frame-frame for implementation. This is like a combination of this scheme with some form of explicit notice will be a full strength to prevent the occurrence of congestion in the Frame Relay network.