Controlling fairness and task granularity in distributed, online, non-clairvoyant workflow executions

摘要

Distributed computing infrastructures are commonly used for scientific computing, and science gatewaysrnprovide complete middleware stacks to allow their transparent exploitation by end users. However, administratingrnsuch systems manually is time consuming and sub-optimal because of the complexity of the executionrnconditions. Algorithms and frameworks aiming at automating system administration must deal with onlinernand non-clairvoyant conditions, where most parameters are unknown and evolve over time. We considerrnthe problem of controlling task granularity and fairness among scientific workflows executed in these conditions.rnWe present two self-managing loops monitoring the fineness, coarseness, and fairness of workflowrnexecutions, comparing these metrics with thresholds extracted from knowledge acquired in previous executionsrnand planning appropriate actions to maintain these metrics to appropriate ranges. Experiments on thernEuropean Grid Infrastructure show that our task granularity control can speed up executions up to a factor ofrn2 and that our fairness control reduces slowdown variability by 3–7 compared with first-come, first-served.rnWe also study the interaction between granularity control and fairness control: our experiments demonstraternthat controlling task granularity degrades fairness but that our fairness control algorithm can compensaternthis degradation.