Optimal buffer management via resource augmentation

We study a buffer management problem in network switches supporting QoS(Quality of Service). Each input or output port has a buffer of limited capacity which can store up to B packets. Packets arrive online and a buffer management policy determines which packets are transmitted through the buffer. After admitted into the buffer by the policy, packets are transmitted in FIFO fashion, that is, in the order they arrive. Each packet has a value and the goal of the policy is to maximize the total value of packets transmitted. The main contribution is to apply resource augmentation analysis to the problem, investigating the optimality of various online policies. The online policy has more resources than the optimal offline policy, that is, additional buffer and a higher transmission rate. For two types of models, nonpreemptive and preemptive, we derive optimal online policies with additional buffer or a higher transmission rate. Also we prove lower bounds of the resources for any optimal online policy.