- Feedback mechanism and
- The control mechanism
Thursday, August 29, 2013
- Traffic shaping is an important part of congestion avoidance mechanism which in turn comes under congestion management.
- If the traffic can be controlled, obviously we would be able to maintain control over the network congestion.
Congestion avoidance scheme can be divided in to the following two parts:
- The feedback mechanism is also known as the network policies and the control mechanism is known as the user policies.
- Of course there are other components also but these two are the most important.
- While analyzing one component it is simply assumed that the other components are operating at optimum levels.
- At the end, it has to be verified whether the combined system is working as expected or not under various types of conditions.
Network policy has got the following three algorithms:
1. Congestion Detection:
- Before information can be sent as the feedback to the network, its load level or the state level must be determined.
- Generally, there can be n number of possible states of the network.
- At a given time the network might be in one of these states.
- Using the congestion detection algorithm, these states can be mapped in to the load levels that are possible.
- There are two possible load levels namely under-load and over-load.
- Under-load means below the knee point and overload occurs above knee point.
- If this function’s k–ary version is taken, it would produce k load levels.
- There are three criteria based up on which the congestion detection function would work. They are link utilization, queue lengths and processor utilization.
2. Feedback Filter:
- After the load level has been determined, it has to be verified that whether or not the state lasts for duration of sufficiently longer time before it is signaled to the users.
- It is in this condition that the feedback of the state is actually useful.
- The duration is long enough to be acted up on.
- On the other hand a state that might change rapidly might create confusion.
- The state passes by the time the users get to know of it.
- Such states misleading feedback.
- A low pass filter function serves the purpose of filtering the desirable states.
3. Feedback Selector:
- After the state has been determined, this information has to be passed to the users so that they may contribute in cutting down the traffic.
- The purpose of the feedback selector function is to identify the users to whom the information has to be sent.
User policy has got the following three algorithms:
- The users to which the feedback signals are sent by the network interpret them after accumulating a number of signals.
- The nature of the network is probabilistic and therefore signals might not be the same.
- According to some signals the network might be under-loaded and according to some other it might be overloaded.
- These signals have to be combined to decide the final action.
- Based up on the percentage, an appropriate weighting function might be applied.
2. Decision Function:
- Once the load level of the network is known to the user, it has to be decided whether or not to increase the load.
- There are two parts of this function: the direction is determined by the first one and the amount is decided by the second one.
- First part is decision function and the second one is increase/ decrease algorithms.
3. Increase/Decrease Algorithm:
- Control forms the major part of the control scheme.
- The control measure to be taken is based up on the feedback obtained.
- It helps in achieving both fairness and efficiency.