Epochs: Trace-Driven Analytical Modeling of Job Execution Times

Queuing theory has extensively studied the problem of estimating job execution times in steady state conditions both in the case of single queues and queuing networks. This paper discusses the use of closed Queuing Network (QN) models during finite time intervals to estimate the execution time of jobs submitted to a computer system. More specifically, the paper presents the Epochs algorithm that allows job traces and randomly generated job inter-arrival times to be used as input to analytic models of computer systems. In other words, the paper combines methods used in discrete event simulation for the characterization of job arrivals with efficient analytic methods to model contention for resources in a computer system. The Epochs algorithm was validated against experimental results using jobs derived from a micro-benchmark and real jobs. The validation shows that the absolute relative error between measurements and execution time predictions obtained with the Epochs algorithm is below 10% in most cases and is at most 15%.