Towards General Distributed Resource Selection

Ming Tai Ha,Matteo Turilli,Andre Merzky,Shantenu Jha

DOI: https://doi.org/10.48550/arXiv.1801.02651

2018-01-08

Distributed, Parallel, and Cluster Computing

Abstract:The advantages of distributing workloads and utilizing multiple distributed resources are now well established. The type and degree of heterogeneity of distributed resources is increasing, and thus determining how to distribute the workloads becomes increasingly difficult, in particular with respect to the selection of suitable resources. We formulate and investigate the resource selection problem in a way that it is agnostic of specific task and resource properties, and which is generalizable to range of metrics. Specifically, we developed a model to describe the requirements of tasks and to estimate the cost of running that task on an arbitrary resource using baseline measurements from a reference machine. We integrated our cost model with the Condor matchmaking algorithm to enable resource selection. Experimental validation of our model shows that it provides execution time estimates with 157-171% error on XSEDE resources and 18-31% on OSG resources. We use the task execution cost model to select resources for a bag-of-tasks of up to 1024 GROMACS MD simulations across the target resources. Experiments show that using the model's estimates reduces the workload's time-to-completion up to ~85% when compared to the random distribution of workload across the same resources.