Some Results on Optimal Decisions in Network Oriented Load Control in Signaling Networks
|Title:||Some Results on Optimal Decisions in Network Oriented Load Control in Signaling Networks|
|Organization:||Blekinge Institute of Technology|
|Department:|| (*** Master error ***)
*** Master Error ***
+46 455 780 00
*** Master error ***
|Abstract:||Congestion control in the signaling system number 7 is a necessity to fulfil the requirements of a telecommunication network that satisfy customers’ requirements on quality of service. Heavy network load is an important source of customer dissatisfaction as congested networks result in deteriorated quality of service. With the introduction of a Congestion Control Mechanism (CCM), that annihilates service sessions with a predicted completion time greater than the maximum allowed completion time for the session, network performance improves dramatically. Annihilation of already delayed sessions let other sessions benefit and increase the useful overall network throughput. This report discuss the importance of customer satisfaction and the relation between congestion in signaling networks and customer dissatisfaction. The advantage of using network profit as a network performance metric is also addressed in this report.
Network profit and network costs are given a stringent definition with respect to customer satisfaction. An expression of the marginal cost for accepting or annihilating sessions is also given. Finally, the CCM is refined using a decision theoretic approach that bases the decision of annihilation on the average profit attached to each of the two possible actions, i.e. annihilate the session or not. The decision theoretic approach use a load dependent probability distribution for the completion time. The results in this report indicate that the decision theoretic approach to the CCM (DCCM) is robust and can handle very high overloads, both transient and focused, keeping the network profit on a high level.
|Subject:||Telecommunications\Overload Control in Intelligent Networks|