Processor-sharing queues: some progress in analysis
Queueing Systems: Theory and Applications
Web server workload characterization: the search for invariants
Proceedings of the 1996 ACM SIGMETRICS international conference on Measurement and modeling of computer systems
Self-similarity in World Wide Web traffic: evidence and possible causes
IEEE/ACM Transactions on Networking (TON)
SIGMETRICS '98/PERFORMANCE '98 Proceedings of the 1998 ACM SIGMETRICS joint international conference on Measurement and modeling of computer systems
Heavy-tailed probability distributions in the World Wide Web
A practical guide to heavy tails
Feedback Queueing Models for Time-Shared Systems
Journal of the ACM (JACM)
On choosing a task assignment policy for a distributed server system
Journal of Parallel and Distributed Computing - Special issue on software support for distributed computing
Two-level processor-sharing scheduling disciplines: mean delay analysis
Proceedings of the joint international conference on Measurement and modeling of computer systems
A least flow-time first load sharing approach for distributed server farm
Journal of Parallel and Distributed Computing
Systems with multiple servers under heavy-tailed workloads
Performance Evaluation - Performance 2005
Task assignment with work-conserving migration
Parallel Computing
ACM SIGMETRICS Performance Evaluation Review
The Foreground-Background queue: A survey
Performance Evaluation
M/G/1/MLPS compared to M/G/1/PS
Operations Research Letters
Hi-index | 0.00 |
Recent research indicates that modern computer workloads (e.g. processing time of web requests) follow heavy-tailed distributions. In a heavy-tailed distribution there are a large number of small tasks and a small number of large tasks. The rationale for using a multi-level time sharing policy is that it can minimise both waiting time and slowdown of tasks that require relatively small service requirements. This in turn will improve the overall performance of the system. Using a 2-level system (policy), we investigate the effect of quanta on the overall performance of a multi-level time sharing policy under a range of workloads and task size variabilities. We measure the performance using slowdown and flow time. First, we show that for most workloads and task size variabilities there exists a unique set of quanta ('optimal' set of quanta) that would result in the best performance. Second, we investigate the performance degradation in one metric under the optimal parameters of other metric. Through an extensive numerical analysis, we find that under high system loads and task size variabilities using the optimal set of quanta corresponding to overall expected slowdown can result in the overall expected flow time to deteriorate significantly. Finally we show that a 3-level system with the optimal set of quanta outperforms a 2-level system with the optimal set of quanta for all the scenarios considered.