计算集群部署模块的设计与实现-软件工程专业论文.docxVIP

哈尔 哈尔滨工业大学工程硕士学位论文 - - II - Abstract In the information era, the logic of business operation is more complicated and the computational data increases tremendously, therefore much higher performance is required for the scientific computing. But it will cost too much to purchase super-computers. In order to reduce such cost, building grid computing clusters by using PCs become first choice for customers. System administrators need unite-vision to deploy the cluster and manage the configuration. The software should monitor the status of the cluster, not only the hardware status but also some parameters that set by administrators themselves. Based on the above requirements, the paper investigates the deployment protocol of cluster deployment, file transfer protocol and cluster performance monitoring issues. We set a system that can be used to deploy cluster, configure and monitor the cluster. Software uses PXE (Preboot execution environment) protocol to install the Operation system and 3rd part softwares on remote node. In order to resolve the head node overloading issue when deploying large-scale cluster, we finger out the concept and algorithm of Maximum deployment node. When the number of to be deployed nodes exceeds Maximum deployment node, Deployment engine will automatically trigger to P2P (peer to peer) protocol to transfer the data. This will protect deployment job failure due to overload on head node. As using BT (Bit-Torrent) to achieve P2P protocol, we also modify the algorithm of transferring on last data blocks, which improves the speed of data transmission. In system monitor module, software can monitor not only the CPU load, CPU usage, memory usage, Cache, Buffer, Swap, but also the disk input/output information, the network load and usage status etc. System admin can use this kind of information to improve the performance of the whole cluster. Keywords: Cluster Deployment, PXE Protocol, BT Protocol - - III - 目 录 摘 要 I Abstract II 第 1 章 绪论 1 1.1