Dynamic Resource and Task Allocation for Time Efficiency in Modular Reconfigurable Space Systems

Abstract

Space unmanned systems with modular and reconfigurable characteristics show a distributed spatial information processing scenario, which needs a new intelligent information processing architecture and task scheduling method, so as to make better use of on orbit resources, adapt to reconfigurable spatial computing and provide spatial reliability computing. we propose a dynamic computing resource and task allocation algorithm for modular reconfigurable space systems, by introducing Docker container technology to minimize completion time cost of a task for given resource constraints considering network traffic and variable workloads in a unified framework. The algorithm first answers the problem of how multiple modules perform cooperative and reliable computing in modular reconfigurable space systems, and gives a method to jointly optimize task allocating strategy and computing resource scheduling through online prediction and linear regression optimization in distributed environment. We design and implement the algorithm architecture on the Linux based platform, and perform series of tests on the simulated distributed network environment. The experiment results show that the algorithm has resource utilization and task response gain in various situations (different cluster node capabilities, different task response demands), obtain the cluster resource allocation balance, and minimize the total completion time of tasks.