Priority-based Task Allocation Strategy in Remote Sensing Image Parallel Processing
FU Zheng-y · Jisuanji gongcheng · 2014
After a discussion of the low efficiency task allocation and load imbalance problem in parallel computing of remote sensing image, this paper gives the multi-task distribution model, and proposes a compute nodes' priority-based task distribution algorithm, which is in comprehensive consideration of load and performance of compute nodes. It collects real-time information on each node when the task is assigned. According to the formula to calculate the priority of each computing node and in accordance with the priority level of assign tasks, under the premise of load balancing, this algorithm can assign task reasonable to computer nodes. Evaluation result shows that this algorithm can process task allocation in reasonable time. The distribution of the tasks becomes more reasonable and uniform, and behaves better than round-robin scheduling algorithm by about 2 times when the number of tasks increases.