A Two-way Matched Dynamic Task Scheduling Algorithm And Its Application On The Industry Clusters Of Remote Sensing Product

With the rise of commercial remote sensing and the integrated use of remote sensing, and the progress of other related technology has made, not only the quality of remote sensing data has been to a new level, but also the volume of remote sensing data has a sharp increase, which challenges the business of processing remote sensing data. In order to enhance the effectiveness of the remote sensing data system, developing an efficient cluster system for improving remote sensing data processing becomes necessary. Task scheduling algorithm as the core component of a high-performance cluster processing system, plays a key role for improving the performance of the whole cluster system during the processing process, which is attended by the research and engineering personnel.The feature of cluster system of remote sensing satellite is that it can handle a lot of task, and its task type is relatively simple. Also the computing node utilization is high balanced and so on. A number of clusters task scheduling algorithm have some deficiencies in some areas, such as computing node utilization is highly uneven, system throughput is small, the average waiting time for task scheduling is long, etc.. It is difficult to meet the needs of remote sensing data processing. To solve these problems, this thesis is on the basis of research on task scheduling algorithm for the remote sensing satellite cluster system. Thesis will give an assessment of remote sensing data processing in two perspectives, one of which is resource utilization, and the other is efficiency of processing. Thesis finally will give an algorithm for remote sensing data processing, with dynamic two-way matching scheduling function. The algorithm has significantly improved the effectiveness of remote sensing data processing cluster system by the experimental results.The main contents and contributions of this thesis:(1) Design and implement a dynamic two-way matching Scheduling AlgorithmBy anglicizing the remote sensing satellite cluster system, thesis gives a task scheduling algorithm model, which combined with remote sensing data processing bidirectional dynamic scheduling algorithm to match a priori values. The scheduling model is made up of task-oriented and model-oriented computing node queue queuing model. In the case of combination of a priori values, consider task-oriented tasks on the one hand the amount of waiting time and tasks to determine the size of the task queue, on the other hand while facing the computing resources in a dynamic computing resources to obtain the load factor, the computing resources to form a queue; the optimal computing resources to match the highest priority tasks to complete the task and matching bidirectional dynamic computing resources.(2) Design and implement a strategy of dynamic switching task scheduling algorithmThere will be new task scheduling algorithm in the optimization process of cluster processing system, but with the different task of cluster processing system and different environments, the new algorithm may not be suitable for all operating scenarios, while in the past some of the scheduling algorithm applies to certain scenarios depending on the cluster system processing tasks or scenarios and switch to a different task scheduling algorithm, by doing that the cluster system can run effectively. This thesis designs and implements a strategy of dynamic switching task scheduling algorithm, which is based on the value of extracted features task scheduling algorithm to find the maximum impact factor eigenvalues, scheduling algorithm switching, increasing the flexibility to cluster system.(3) Thesis will propose a combination of remote sensing data processing priori with dynamic scheduling algorithm and a matching bidirectional design task scheduling algorithm and apply to remote sensing satellite cluster production system. After the adoption of a large number of experiments, the experimental results show that the task scheduling algorithm can improve the performance of the proposed system-based group, namely improving the utilization of computing node, shortening the processing time of remote sensing data to ensure the reliability of the cluster system.