Security Management And Energy Saving Control Of Parallel Cluster

Because of low cost, simple structure, easy use and good scalability, cost performance of cluster systems is much higher than large-scale dedicated computer system.Experts predict that:the future of high-performance computers and super-server will be based on cluster structure.Near the planet’s magnetic field simulation, to predict the emergence of a tornado, location of oil resources, etc. are stored in large amounts of data need to be processed.The traditional approach is to use a super computer to do computational work.However, super-computer is more expensive; the availability and scalability are not strong enough.Therefore, high-performance computing clusters become the focus of attention.This paper expounds the importance of parallel computing, the developing situation of parallel clusters, and parallel computer cluster nodes environment of hardware and software in the laboratory in detail.At present, cluster nodes still faces many problems and management.For these problems, this thesis has analyzed them, and proposed the feasible solutions. This scheme mainly on the basis of the fleet system is carried out from two aspects of hardware and programming. Under the support of the clusters system software,Paper mainly studied that working nodes in the control of server node and saving energy control.Main content of this thesis is that service nodes can get the state node information of the each work nodes in the working process through programming and network communication.And these state data can be instantly display and updated on the user management interface in the server node.Then according to these state information.working nodes are reasonably managed and controlled. Working node status information mainly contains of nodes numbers,CPU usage, CPU chip temperature,memory usage,hard disk usage,contion of the network,etc. Working nodes run in Linux operating systems, So it can gain needed data by the dynamic hardware for measurement file system pro of Linux operating system.Through the network programming working node status information will be transmitted to service node and displayed with GTK,in order to promptly observe each participating node of the current hardware computing resource utilization, and to close working nodes in an emergency with SSH avoiding accidents. Because the software realization exists some limitations, this paper also describes methods of the hardware system of management and control in theory.