Cluster System Load Balancing Research And Implementation Based On Linux OS

With in-depth development of high-performance computing and scientific computing, cluster system has been received attention increasingly. Cluster computing system with low-cost, high-performance features, including powerful batch processing and parallel computing capacity, represents the mainstream development of the high-performance computer. How to improve the operating efficiency, availability and easy to use of the cluster system is an important issue in computing field. In such systems, improving hardware performance can not fully meet diverse requirements of users and particularly can not meet the requirements of mass scientific computing and commercial applications, so it is necessary to manage the computer resources effectively.Being an important component of the cluster software system, the major design goal of the cluster management system is to achieve the above requirements. Although with a perfect capacity of management, cluster management system popularly can't schedule resources of hardware and software dynamically based on different application, also its management and adaptive capacity is yet to be further improved. In this paper, we study the load balancing scheduling algorithm and resource management optimization technique systematically. After analyzing the model structure of the cluster system, we improve the original algorithm and solve the resource-scheduling problem greatly. At the same time, a Linux-based cluster system terminal node dynamic database backup system is designed and realized. Other than the capacity of load balancing scheduling of resources dynamically inside the cluster system, our cluster management system is also able to perform well in load balancing scheduling among different cluster system after the introduction of new algorithm. Experiments show that our system outperforms traditional system in terms of throughput and average respond delay.