Service Curve Allocation For End-to-end Delay Requirement

Abstract

An application with guaranteed service only cares about whether or not the network can satisfy its performance requirement, such as end-to-end delay. However, the network wants to achieve the high utilization and performance guaranteed simultaneously. If the end-to-end delay provided by the network can be allocated properly to each switching node, then the network resources can get a better utilization. Conventionally, the delay is allocated equally to each switching node along the path that the connections pass through, referred to as even division policy. The advantage of this policy is easy to implement. However, we can not understand how this policy will affect the network utilization. In this paper, we proposed an allocation scheme called MaxMin allocation to improve network utilization. Under SCED (Service Curve Earlist Deadline) scheduling policy, we reduce the service curve allocation to end-to-end delay allocation, and using MaxMin delay allocation policy, to promote the network utilization. From simulations, we find that the performance of MaxMin policy is better than even division policy.