Computing Missing Service Demand Parameters for Performance Models.

Abstract

One of the challenges in building analytic performance models such as queuing network models is obtaining service demands for the various workloads and various devices. While some of these parameters can be easily measured, some may not be easy to obtain due to the complexity that the measurements may entail or because it may not be possible to stop the operation of a production system to collect measurements. This paper discusses a black-box approach for computing unknown service demand parameters in queuing network models. The paper addresses the problem of finding a subset of the service demand values given the known values and given the values of the response times for all workloads. A unique closed form solution is given for the case of a single missing parameter and a process for obtaining a feasible solution for the case of multiple missing parameters is discussed. Numerical examples illustrate the approach. An online service demand estimator that successively computes better estimates for the service demands is described. The experiments carried out with this estimator show a relatively low relative error in the predicted response times when the estimated service demands are used. 1

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 40 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.