網(wǎng)絡(luò)工作站以其低價高效的計算環(huán)境和快速可靠的服務(wù)被人們廣泛接受并應(yīng)用于高
性能計算。按其用戶使用類型可把網(wǎng)絡(luò)工作站分為專用網(wǎng)絡(luò)和非專用網(wǎng)絡(luò)，專用工作站具有相對結(jié)構(gòu)和管理簡單，易于擴展等特點，非專用工作站具有充分利用計算機資源的特點，但管理較復(fù)雜。文中提出一種非專用網(wǎng)絡(luò)計算的性能預(yù)測數(shù)學(xué)模型，它包括單個工作站系統(tǒng)的利用和同構(gòu)機群系統(tǒng)的分布式服務(wù)。通過這個計算模型可從機器利用率，連續(xù)工作服務(wù)速率，并行任務(wù)分配等方面對并行任務(wù)完成時間進行預(yù)測。它為非專用網(wǎng)絡(luò)環(huán)境的任務(wù)調(diào)度提供有價值的參考。
關(guān)鍵詞：性能預(yù)測模型；集群計算；非專用網(wǎng)絡(luò)系統(tǒng)；負載分布
abstract: The low cost and wide availability of networks of workstations have made them an attractive solution for high performance computing. According to the ownership, NOW can be divided into dedicated networks of workstations and non-dedicated networks of workstations. The structure and management are easier in dedicated NOW and is prone to expand. The non-dedicated NOW can make great use of computing resource, but the
administration is not easy. A non-dedicated NOW computing performance forecast model is advanced in this paper. There are two parts―one workstation system’s utilize and the distributed service of homogeneous cluster. The computing model result in the separation of the influence of machine utilization, sequential job service rate, and parallel task allocation to the parallel task completion time. It is a useful reference for non-dedicated network
environment.
Key word: performance forecast model, cluster system, non-dedicated systems, workload distribution.