应用决策集选播模式的SIP测控网络负载均衡策略

针对目前单优化目标选播模式容易导致负载发生新的不均衡问题,建立了SIP测控网络平台并分析其选播机理。采用网络层选播和应用层选播综合的方式,提出了最优决策集选播的负载均衡策略。选播目标为一个集合,包含多个满足要求的服务器,以避免蜂拥抢占服务器资源的情况发生;建立SIP测控网络最优决策集选播数学模型,采用证据理论决策选播目标权重,从而产生最优决策集。在OPNET仿真环境下,搭建SIP测控网络平台进行实验,结果表明最优决策集选播模式比单优化目标选播模式的服务器占用资源不均衡率降低了31%,能够很好地将负载均衡到各个服务器中,并减少了单个服务器处理多进程需要的额外资源开销。

Load-balancing strategy in SIP measuring & controlling network based on anycast mode of decision set

To solve the problem that a single optimal target anycast mode tends to cause new load imbalance, a Measuring & Controlling Network Platform(MCNP) based on the Session Initiation Protocol(SIP) was established and its anycast mechanism was analyzed. By integrating network-layer anycast and application-layer anycast, an load-balancing method based on an optimal decision set was introduced.The anycast objective is a set,which includes many servers to avoid server resources to be seized confusedly. To obtain an optimal decision set, a mathematical model of optimal decision set anycast was established,and the evidence theory was used to anycast the weight. An experimental platform was built to simulate and verify this new method in OPNET simulation environment. The results show that the imbalance rate of server resource occupancy by using the optimal decision set anycast mode has reduced by 31% as compared with that by using a single optimal target anycast mode, which brings an equal distribution of load to different servers and reduces the extra resources for tackling multi-processing in a single individual server.