遥感影像处理特征及数据密集型任务调度算法研究

Research of Character Analyzing and Data Intensive Task Dispatching in Remote Sensing Image Process

针对云计算技术在遥感影像处理中的应用问题,对于其任务特征以及调度问题进行了研究。首先定义了遥感影像处理工作流,并在特征分析的基础上对任务类型进行了划分,将任务划分为数据密集型和计算密集型两种。针对数据密集型任务调度问题,通过建立流程间的数据依赖度矩阵和数据集划分制定调度策略,设计了以减少数据传输次数为目标的调度算法。实验结果表明,该算法与随机算法相比,可以有效减少数据集、数据服务器数量和不同比例的固定数据集等多种情况下的数据传输次数,该算法能够满足常用数据密集型任务的调度需求,如计算航段信息、计算缩略图同名像点等。

According to the problem of cloud computing technology applying in remote sensing image process,the task characters and scheduling problem is researched in this paper. Firstly,the remote sensing process workflow is defined,then the tasks are divided into data intensive task and compute intensive task based on the task characters analyzing. According to the problem of data intensive task scheduling problem,an algorithm was proposed. The data dependency matrix is built among different process and then the dataset is divided to make the scheduling strategy. The scheduling objective is reducing the number of data movement in the computing. Experiments illustrated that compared with random algorithm,the algorithm could effectively reduce the number of data movement in different experiment conditions like various dataset,data service number and fixed dataset,which could solve the data intensive task scheduling problem in remote sensing image process like the airline information computing and the thumbnail corresponding point extraction.