Construction And Management Of Parallel Computing Cluster

Parallel computing technology at home and abroad are a high priority, it has made great achievements in scientific research, engineering technology and the application of military technology. Scientists widely believe that the international 21st century the important technology issues facing the enormous challenge known as the human genome, to accurately predict the global climate, ocean circulation loop, etc., not trillions of times more than the high-performance computers can not solve. Simulated nuclear explosions on the military must use more than one trillion high-performance computers.Parallel computing applications in today's increasingly complex problems to solve important means. In this paper, cluster systems, the cluster system performance monitoring and performance tuning and so a systematic introduction to the relevant application of parallel computer cluster technology. The main innovative contributions include:1.Discussed the better way to build a cluster system and involved in the construction of the relevant hardware and software.2. Designed to aid the management of a semi-automated shell script. The script is the main identification with the operating system, user management, document management, shared management, service management, in the cluster to synchronize the command, synchronization of all nodes in the cluster system time and test network connectivity and other features.3. Discussed the problem of cluster performance monitoring system, on the CPU, memory, IO and network sub systems.4. To find a series of clusters useful ways for tuning the program, can effectively optimize the computing nodes subsystems, network transmission subsystems and cluster IO subsystem's performance.5. Analysis of the various sectors of the cluster system software application, the parallel capacity on the part of the software made certain experimental analysis. This paper studies the parallel computer cluster based on Linux building, cluster performance testing and tuning, and semi-automatic fleet management. Construction of a parallel computer cluster involved in key equipment, software and hardware components of the cluster and the cluster-building steps are described in general. Shell script programming through a semi-automated fleet auxiliary management tools, you can achieve the cluster in the user management, document management, shared storage management, system management and network connectivity services, testing and other functions. Cluster monitoring of the implementation of the five technical way, focusing on through the/ proc virtual file system to achieve the principle of cluster monitoring, and through C language programming on/proc directory associated with the system running data collection. Finally, the parallel computer cluster applications in various fields, and involved in all areas of professional application software features were introduced.The semi-automatic paper has been applied to practical project management tools, optimization of the fleet has also been tested in the real cluster systems, significantly improving the manageability of cluster and give full play to the cluster's overall performance.