Research And Application Of Task Scheduling Model In Massive Remote Sensing Image Common Product Generation

Along with the huge demand for information data, there is an explosive growth of remote sensingimage data. And the huge demand for data processing in remote sensing application also greatly promotesthe highly development of remote sensing image processing technology. High performance computingcluster is one of the important technologies to implement the rapid processing of massive high resolutionsatellite remote sensing data, and it can effectively alleviate the bottleneck of high resolution remotesensing application. Among them, the task scheduling model is the key to improve the efficiency of clustersystem.The massive remote sensing image common production subsystem of the country’s majorprojects subject high-resolution earth observation application system which the author participate, has thecharacteristics of single nature task, light weight sub-task, large resource file and relatively independent, etc.The traditional task scheduling model is poor targeted, poor practicability, and its scheduling time is toolong, it can’t meet the requirements of the system. In order to solve the above problems, this paper adoptsmulti-index load balancing technology and hierarchical task scheduling methods to improve the existingscheduling model, and put forward a kind of two-level task scheduling model combined with loadbalancing. The main work and contributions are as follows:1. Based on the analysis of massive remote sensing image common production system and thescheduling strategy of current task system, this paper proposes a two-level task scheduling modelconsidering the load balance. The model is aiming at the questions of small throughput, poor reliability,poor fault tolerance etc. 2. Based on the model, the author designs the load balancer and the interface between taskmanager and the load balancer through the analysis of load balance technology at first. Secondly, accordingto the characteristics of the massive remote sensing products production system, this paper accomplishesthe design of the whole processing task scheduling model according to the characteristic of the task system.Finally, designs the first-order and second-order task scheduling model.3. Applying the improved task scheduling model to the massive remote sensing image commonproduct generation system, and through a large number of experimental data validation, the improved taskscheduling model was better than the existing model. In the respects of the system throughput, adaptability,reliability, fault tolerance, etc., the efficiency is greatly improved. What’s more, the improved system evenhalved the processing time of certain satellite data slicing.