Shanxi Normal University High Performance Computing Cluster Design And Performance Optimize

With development of the computer cluster technology, especially the high-performance computer cluster technology, the large-scale massive data processing issue has gradually become the matter to which scientific and technical personnel pay the most attention. Shanxi Normal University High performance computer cluster structure design and research have been a hot issue; rapid development of modern server hardware technology and acceleration in network communication provide powerful support for computer cluster development. In the context of this technology, this paper gives an idea based on server stack technology and preliminarily completes computer cluster structure design, implementation, and verification by following advanced OS system design and tools in combination with current massive computing demand. For this issue, this paper offers a computer cluster structure method based on multi-node stack technology by combining the current characteristics of massive computation, and on this basis, carries out research and design.The following is an outline of the main research content and results of this paper.It first gives a systematic summary and explanation of the basic knowledge about the computer cluster infrastructure, analyzing the computer cluster features and functions and pointing out that combination of server hardware technology with network transmission technology is an important way to solve the current issue. On this basis, this paper constructs a systematic computer cluster model and analyzes and studies all parts of this model, pointing out the most important inter-node communication and choice of multi-node management in a computer cluster. It also analyzes and studies the current mainstream computer cluster structure and the characteristics of the computer cluster hardware infrastructure and OS, with computer cluster systems compared. This paper gives cluster structure to be constructed by using PCserver and the Linux operating system.The dual-network interface card (NIC) binding technology achieves load balancing and fail safe, and the NIC binding technology under Linux improves reliability of servers and increases available network bandwidth, furnishing users with uninterrupted critical services. To maintain high availability of servers is an important factor of the IT environment for massive computing users, and one of the most important things is high availability of a server's network connectivity. The NIC binding technology helps to ensure the characteristic of high availability and provides other advantages to improve the network performance.For remote access Shell. Facilitate instructions to be executed on a specified remote host computer. Open the rsh, rcp, and rlogin command authority. Set up cross-server access and mutual trust, so that rsh and rcp can be executed without needing a password.The NFS network file system allows sharing of other network computers with designated folders in the system by using NFS. We can often use NFS to expand space of the system and to realize resource sharing. For NFS configuration-related issues, it is pointed out that NFS is a key part of the computer cluster file system through NFS analysis and research.With the NIS (network information service) to management work, it is allowed to use management files in the central server. NIS is a client/server system, and ypbind is the client process to define a NIS server. Once the server location is determined, the client is bound to the server, so enquiries into the client's address are sent to the server. ypserv is the server process to answer client enquiries. Regarding unified cluster management procedures, service management provides all node services and their state enquiries and can simultaneously carry out start and stop operation on the same service on multiple nodes. A node information module allows view of node file system information, hardware information, NIC configuration information, and system configuration information. Process management is mainly for viewing node process information and executing suspension, recovery, and termination operation on these processes according to process options.