Cell loss analysis for some alternative priority queues

摘要

The paper proposes and analyzes two alternative priority multiplexing schemes together with the associated Call Admission Control (CAC) algorithms. In one of the proposals we assume the well-known, two queue, strict-priority multiplexer with a buffering resource of fixed storage capacity that is divided into two disjoint parts representing the first and second priority queues. In traditional architectures the size of the first and second queues can be set off-line only during configuration procedures. The drawback of such a static configuration is the inability to adapt to the dynamic changes of traffic demands, which in turn may result in an inefficient utilization of both buffering and server resources. By using Large Deviation analysis, we show that there exists an optimal buffer sharing policy that distributes the total available buffer space between the first and second queues according to the current input traffic such that the attainable server utilization will be the highest. Our proposed dynamic mechanism relies on online traffic measurements in making the CAC decision and reconfiguring the buffers at the same time. As another way of improvement we analyze a new multiplexing scheme that includes a scheduling and buffer management algorithm, which is intended to be an efficient way of multiplexing delay but not loss-sensitive voice with loss but not delay sensitive data. In this setup we have only one joint buffer for both voice and data packets. Voice packets have service priority and data packets have storage priority, which means that an incoming data packer can push-out a voice packet from the buffer if there is no free space available. We also analyze this multiplexer, propose an associated CAC algorithm and compare it with the previous one.