Research Of Multi-domain And Multi-overlay Mechanisms For Innovative Mobile Service Control Network

IMS (IP Multimedia Subsystem) as the focus of next generation telecommunicationnetwork has been actual deployed by domestic and foreign operators. While the originalIMS adopts the traditional centralized architecture, which leads to poor scalability, highdeployment cost, and load unbalance. So introducing P2P technology into IMS corenetwork has become a feasible solution to those problems. While the existed solutionscombine different network functions into one function in only one overlay, which againstthe evolution tendency of separation of IMS network element function and couldn’t meetthe requirements for deploying in different domains. In order to ensure the characteristicsof the home domain and network elements’ separation in original IMS, we propose themulti-domain and multi-overlay mechanism for core network.To ensure the speration of funtions in core network, we use the achievements of IETFP2P SIP work station RELOAD (REsource LOcation And Discovery) protocol forbuilding core network overlay, then elements join the overlay with different identities.With using RELOAD, we propose the architecture of multi-domain and multi-overlay tosatisfy the request of deploying IMS core network with domain. Then, in order to ensurethe data resources reliably stored in P2P overlay, we propose the algorithms for data’spublishing, data copies’ maintenance and searching. According to the situation that nodesmay join or leave the core network overlay dynamically, we propose the methods fordiscovering nodes by querying DNS and address list published in overlay network, andimplement the strategies for load balance. At last, we optimezed the procedures forcommunications to take advantages of overlay network to achieve distributed sessionmanagement and ensure the time for handling sessions.The whole system is deploied in Hainan branch of China Unicom. With large scale oftesting, the scalability of core network is proved, the load can be distributed to every nodebalanced and the difference of load is less than16%, and the optimezed procedures canensure the speed for session handling with little decreased compared with originalprocedures.