An Investigation into Intelligent Network Congestion Control Strategies

Abstract This thesis examines the congestion control issues that arise in Intelligent Networks, when it is necessary to support multiple service types with different load requirements and priorities. The area of Intelligent Network (IN) congestion control has been under investigation for over a decade, but in general, the models used in this research were over-simplified and all service types were assumed to have the same priority levels and load requirements at the various IN physical elements. However, as the IN is a dynamic network that must process many different service types that have radically different call load profiles and are based on different service level agreements and charging schemes, the validity of the above assumptions is questionable. The aim of this work, therefore, is to remove a number of the classic assumptions made in IN congestion control research, by: • developing a detailed model of an IN, catering for multiple traffic types, • using this model to establish the shortcomings of classic congestion control strategies, • devising a new IN congestion control strategy and verifying its superiority on the model. To achieve these aims, an IN model (both simulation and analytic) is developed to reflect the physical and functional architecture of the network and model the information flows required between network entities in order to execute services. The effectiveness of various classic active and reactive congestion control strategies are then investigated using this model and it is established that none of these strategies are capable of protecting both the Service Control Point and Service Switching Points under all possible traffic mixes and loads. This is partially due to the fact that all of these strategies are based on the use of fixed parameters (and are therefore not flexible enough to deal with IN traffic) and partially because none of these strategies take into account the different load requirements of the different service types. A new, flexible strategy is then devised to facilitate global IN congestion control and cater for service types with different characteristics. This strategy maximises IN performance by protecting all network elements from overload while maximising network revenue and preserving fairness between service types during overload. A number of factors determining the relative importance or weight of different traffic types are also identified and used by the strategy to maintain call importance during overload. The efficiency of this strategy is demonstrated by comparing its operation to that of the best classic IN overload controls and also to a new strategy, which has scalable and dynamic behaviour (and which was devised for the purpose of providing a fair comparison to the optimisation strategy). The optimisation-based strategy and dynamic strategy are found to be equally effective and far superior to the classic strategies. However, the optimisation algorithm also preserves relative importance and fairness, while maximising network revenue but at the cost of a not insignificant processing overhead.This thesis examines the congestion control issues that arise in Intelligent Networks, when it is necessary to support multiple service types with different load requirements and priorities. The area of Intelligent Network (IN) congestion control has been under investigation for over a decade, but in general, the models used in this research were over-simplified and all service types were assumed to have the same priority levels and load requirements at the various IN physical elements. However, as the IN is a dynamic network that must process many different service types that have radically different call load profiles and are based on different service level agreements and charging schemes, the validity of the above assumptions is questionable. The aim of this work, therefore, is to remove a number of the classic assumptions made in IN congestion control research, by: • developing a detailed model of an IN, catering for multiple traffic types, • using this model to establish the shortcomings of classic congestion control strategies, • devising a new IN congestion control strategy and verifying its superiority on the model. To achieve these aims, an IN model (both simulation and analytic) is developed to reflect the physical and functional architecture of the network and model the information flows required between network entities in order to execute services. The effectiveness of various classic active and reactive congestion control strategies are then investigated using this model and it is established that none of these strategies are capable of protecting both the Service Control Point and Service Switching Points under all possible traffic mixes and loads. This is partially due to the fact that all of these strategies are based on the use of fixed parameters (and are therefore not flexible enough to deal with IN traffic) and partially because none of these strategies take into account the different load requirements of the different service types. A new, flexible strategy is then devised to facilitate global IN congestion control and cater for service types with different characteristics. This strategy maximises IN performance by protecting all network elements from overload while maximising network revenue and preserving fairness between service types during overload. A number of factors determining the relative importance or weight of different traffic types are also identified and used by the strategy to maintain call importance during overload. The efficiency of this strategy is demonstrated by comparing its operation to that of the best classic IN overload controls and also to a new strategy, which has scalable and dynamic behaviour (and which was devised for the purpose of providing a fair comparison to the optimisation strategy). The optimisation-based strategy and dynamic strategy are found to be equally effective and far superior to the classic strategies. However, the optimisation algorithm also preserves relative importance and fairness, while maximising network revenue but at the cost of a not insignificant processing overhead.