Utility Driven Dynamic Resource Management in an Oversubscribed Energy-Constrained Heterogeneous System

Abstract

Abstract-In this paper, we address the problem of scheduling dynamically-arriving tasks to machines in an oversubscribed heterogeneous computing environment. Each task has a monotonically decreasing utility function associated with it that represents the utility (or value) based on the task's completion time. Our system model is designed based on the environments of interest to the Extreme Scale Systems Center at Oak Ridge National Laboratory. The goal of our scheduler is to maximize the total utility earned from task completions while satisfying an energy constraint. We design an energy-aware heuristic and compare its performance to heuristics from the literature. We also design an energy filtering technique for this environment that is used in conjunction with the heuristics. The filtering technique adapts to the energy remaining in the system and estimates a fair-share of energy that a task's execution can consume. The filtering technique improves the performance of all the heuristics and distributes the consumption of energy throughout the day. Based on our analysis, we recommend the level of filtering to maximize the performance of scheduling techniques in an oversubscribed environment.

Towards Energy Budget Control in HPC

“…While this approach is interesting, the algorithm they used is not scalable and is hardly extendable with other constraints. In the study [20], Khemka et al maximize a "utility" function in a cluster with daily energy budget. They solve the problem thanks to an offline heuristic.…”

Stochastic-based robust dynamic resource allocation for independent tasks in a heterogeneous computing system

“…Dropping tasks is a common practice in oversubscribed systems that have a real time or near real time nature (e.g., [19,20,21]). In these systems, typically, there is no value in executing a task that has missed its deadline.…”