Call admission and efficient allocation for delay guarantees

ATM networks are connection-oriented. Making a call requires first sending a message to do an admission control to guarantee the connections' QoS (qua lity of service) in the network. In this paper, we focus on the problem of translating a global QoS requirement into a set of local QoS requirements i n ATM networks. Usually, an end-user is only concerned with the QoS require ments on end-to-end basis and does not care about the local switching node QoS. Most of recent research efforts only focus on worst-case end-to-end de lay bound but pay no attention to the problem of distributing the end-to-en d delay bound to local switching node. After admission control, when the ne w connection is admitted to enter the network, they equally allocate the ex cess delay and reserve the same bandwidth at each switch along the path. Bu t, this can not improve network utilization efficiently. It motivates us to design a novel local QoS requirement allocation scheme to get better perfo rmance. Using the number of maximum supportable connections as the performa nce index, we derive an optimal delay allocation (OPT) policy. In addition, we also proposed an analysis model to evaluate the proposed allocation sch eme and equal allocation (EQ) scheme in a series of switching nodes with th e Rate-controlled scheduling architecture, including a traffic shaper and a non-preemptive earliest-deadline-first scheduler. From the numerical resul ts, we have shown the importance of allocation policy and explored the fact ors that affect the performance index.