中文摘要：

该文针对多进程共享处理机资源的软交换实体,提出基于非强占、多优先级消息排队的M/G/1/n排队网络性能分析模型.该模型中消息的处理服从定长分布而不是泊松分布,并且存在呼叫损失,更加接近实际系统.给出了消息平均排队时间的解析表达式,理论分析与仿真结果表明相对于无优先级M/G/1/n排队模型,上述模型具有更大系统吞吐量,更高CPU有效负荷,但呼叫接续时间稍有增加.同时分析了消息缓冲区n对于系统性能的影响.

英文摘要：

One M/G/1/n queuing network model based on non-preemptive multi-priority messages is proposed for the performance analysis of softswitch entity, which contained multiple processes sharing the same CPU resources. In this model, processing time of messages is drawn from uniform distribution but not exponential distribution and messages exceeding the queue length threshold are discarded orderly according to their priorities. A formula for the mean time of message processing is derived. Academic analysis and simulation results validate that M/G/1/n model with multi-priority messages has larger throughput and higher effective CPU payload than M/G/1/n model with non-priority messages, but has a little longer message processing procedure. Simulation results also indicate that message buffer has a considerable influence on the performance of queuing system.