Application of Dynamic Allocation Task in Multi Processor Platform

摘要

Traditional multi GPU processor cluster task scheduling and allocation method uses the time gap allocation method does not consider each task execution requirements and each computer node of actual operation ability, the overall efficiency of the implementation. Proposed a hybrid scheduling multi GPU processor platform dynamic task allocation method, using multi processor simulation of multiple computing nodes in a computer system, for different tasks, task computation demand and allocation according to the computer's actual operation ability, the hybrid task scheduling, the system has good robustness and experimental tests were conducted using a set of practical tasks. The results show that using mixed task, and realize the dynamic task allocation method, task execution time is greatly reduced, with good efficiency improvement results.