Research and implementation on an admission control policy for temporally correlated workloads in virtual application environments

It is a challenge to decide when to discard and how to discard requests to ensure the application´s performance to the extreme extent and the ratio of discarding to the least extent in virtual application environments (VAEs) in face of the heavy workloads with temporal correlation characteristic. In this paper, we adopt the admission control policy with the consideration of temporal correlation of workloads under the overload condition. When the request arrives, the policy will check the length of waiting queue. If the queue´s length is over the threshold, the policy will discard or accept the request according to the autocorrelation coefficient of the arriving process or the service process. We call this policy the autocorrelation-based admission control policy (abbreviated as ACF). Our experiments show that ACF is more effective than the traditional random discarding policy (abbreviated as Random). Furthermore, the more temporally correlated the arriving process is, the more effective the ACF is.