Cost sensitive pushout policy for the buffer management in versatile traffic switch

We are presenting here a buffer management policy for versatile traffic. In the case of ATM switch, we have to drop only one packet of low priority to accommodate a packet of high priority, because all packets (cells) are of same size. Thus if we have to improve QoS of high priority packets we can give the worse treatment to low priority packets. This is a procedure to increase total weighted throughput. But in the case of Differentiated Service model where versatility in data traffic is there due to packets of multiple priorities and different lengths, we may have to drop more than one packet of low priority to accommodate a high priority packet. But one interesting question remains, how many packets of low priority should be dropped to save a single high priority packet? In this paper we have demonstrated that merely treating low priority up to worst extent to improve QoS of high priority doesn't improve the total weighted throughput in DiffServ. model. A line must be drawn beyond which worse treatment to low priority packet must be stopped when the goal is to minimize the probability of loss of total weighted cost. We are presenting here the Cost Sensitive Pushout Policy, which gives a class of buffer management policy to maximize total weighted throughput.