Multivariate Estimation of Resource Utilization Bounds of any Variable Schedule in a Computing System

Abstract

Estimations of resource availability and utilizations in a computer system are necessary to achieve fair allocation of resources and to measure performance of a system. The software system designers often look for computing a limiting bound of CPU-utilizations given a schedule in a concurrent multitasking system, where tasks may have different CPU-affinity and IO-affinity. In complex software systems, the variations of CPU-utilizations due to variable scheduling quanta of the concurrent tasks are difficult to estimate through global time-averaging. This paper proposes a computing model of multivariate functional estimation of limiting bound of CPU-utilizations in a concurrent multitasking system comprised of heterogeneous tasks. An analytical model is formulated to compute dynamics of variable scheduling quanta and CPU-utilizations. The relation between continuous single-variable estimation and sampled multi-variable estimation is established. The integral remainder terms along with values of converging polynomials denoting estimation errors are computed. © Springer International Publishing Switzerland 2014.