Research On Load Balancing Algorithm Of Distributed Service Logic Execution Environment In Next Generation Network

Next Generation Network (NGN) integrates the fixed, mobile and package network, provides abstract network capability through open API, realizes the separation of network operating and service provisioning, and is a service-driven network. The major characters of NGN are openness, distribution, and convergence. The services in NGN are mainly hosted on Application Server (AS). Therefore, the design and the performance of Service Logic Execution Environment (SLEE), which is the core entity of AS, play a major role in service provisioning.However, the research on AS is relatively rare. In this dissertation, the core entity of AS, i.e. SLEE, is studied, including the distributed design of SLEE and its load balancing mechanisms. The main contributions are summarized as below:(1) According to the characters of service in NGN, and the distribution of every functionality of SLEE, a design of two-container based SLEE (TCB-SLEE) is introduced, which provides application container to implement the functions related to service management, and provides instance container to implement the functions related to service execution. Further more, two implementation of TCB-SLEE are introduced and analyzed on the performance and suitable environment.(2) According to the characters of service correlation due to the openness and reuse of service capability, as well as the heterogeneity of SLEE, a multi-service deployment optimization algorithm (MSDOA) is introduced, which is based on the concept of "divide and conquer". Simulation tests prove that MSDOA effectively improves the system load balancing ratio.(3) Based on the algorithm introduced in (2) , a generic algorithm named multi-dimensional load-balancing optimization algorithm for service deployment (MLOASD) is introduced, which is suitable for multiple kinds of resources.(4) Aiming at the herd effect due to update interval in distributed clustering system, three dynamic load balacing algorithm, i.e., AK, WK and their combination AWK, are proposed, which are based on the k-subset algorithm in homogeneous clustering system, and improve k-subset with the concept of load accumulation and node weight. Experientes suggests the three algorithms are suitable for heterogeneous clustering system.(5) According to the results in (4) , the impact on the performance of dynamic load balancing algorithm due to different update intervalss and request arrival rate is studied. Some important conclusions are also drawn.