We presented a performance evaluation study that shows the presence of autocorrelated flows in a multi-tiered system with a closed-loop structure and their performance effects. Via queuing models we have shown that identifying autocorrelated flows in such a system is critical for capacity planning. If autocorrelated flows are ignored, then throughput and utilization of specific devices - metrics often used in capacity planning and admission control - may give a distorted view of system load. For more details, we direct the reader to [2]. Future work on the theoretical side will be to develop analytic models of closed systems that consider autocorrelation. Such mod- els could be used to identify the conditions in a closed system that trigger autocorrelation, as well as the conditions that augment or reduce autocorrelation. On the systems side, we are working on new resource management techniques that take autocorrelation into account. We also intend to widen the set of workloads considered and understand in more detail how autocorrelation arises in the service process.
展开▼
