Research On QoS Of Real-time VBR Service With Shared Cache

With the development of modern communication networks, there are various kindsof services are supported by broadband networks, which include various real-timeservices besides best-effort data services. But the QoS requirements of variousreal-time services are quite different. Therefore, the performance study on integratedservices system based on real-time services has become the focus at present.This paper establishes broadband real-time VBR service system model based on ashared cache. Under the complete sharing strategy on resources, it obtains theprobability distribution of the number of online connections. A variety of real-timeservices’ expressions of call blocking probability is obtained by peak method. Basedon the assumption that each channel can only transmit a fixed-length packet in a unitof time, the packets’ arrival process is analyzed and the probability distribution of thenumber of arrival packets is obtained per unit of time. Meanwhile it analyzes the datapackets’ queuing-service process with multiple ON-OFF sources queue model withshared cache in system. In accordance with the rules of queuing, distribution ofone-step transition probability is obtained and then it gets the steady-state probabilitydistribution of the system, which means the probability distribution of the number ofpackets in the system. With the above as base, with invariance of the system’smaximum servicing rate and the limitation of buffer capacity in consideration, theformulations of packet loss rate, average delay and average throughput are obtained.Compared with the non-caching system, the packet loss rate has declined while theaverage delay has increased by numerical calculation. Also it draws a conclution thatthe packet loss rate, the average delay and the average throughput are not monotonicfunction of real-time service load of some type.In order to guarantee QoS of various types of real-time VBR service and improvethe utilization of network resources, this paper presents two call admission controlstrategies. For the system with shared cache, by studying the influence of the numberof users access network and cache size on the system and simultaneously taking intoaccount the threshold requirements for real-time VBR service, this paper puts forwardtwo CAC strategies that give consideration to both packet-level performanceparameters (packet loss rate and average delay) and connection-level performanceparameters (call blocking probability), which separately minimizes average delay and packet loss rate. Through analysis of monotonic relationship among parameters, thecorrectness of the two strategies is validated. Finally, this paper presents a dynamicCAC strategy, which is applied to the applications of service with a relatively smalldelay but a high require for packet loss rate under the precondition to satisfy callblocking probability and packet loss rate requirements.