近年來全球資訊產業的蓬勃發展,使得目前個人電腦的運算處理能力大幅提昇,直追高性能的工作站。另一方面,隨著網際網路的盛行,電腦網路相關技術之進展也成果非凡,其中高速寬頻網路的問世,能提供使用者更多元且豐富的服務。因此,透過高速網路連接個人電腦來達成高效能計算 (HPC) 的可行性大為提高,且具有計算成本低廉,元件取得便利,系統維護容易…等優點。 本研究所提出的以代理人為基礎之分散式平行運算平台,乃建構在目前廣受歡迎的 Linux 作業系統上,並結合個人電腦叢集 (PC Cluster) 形成一強大的高效能計算資源。本系統的使用者可透過 Web 瀏覽器於「任何時間,任何地點」來存取和執行平行運算服務。代理人負責監督叢集中各電腦的負載狀態,並回報給平行運算伺服器,伺服器根據此負載資訊來分派工作,以協助使用者在最短的時間內完成工作,取得結果。最後經由實驗來評估本系統的加速度 (speed-up) 和效率 (efficiency),以驗證本研究之可行性。
PCs and their peripherals are probably the most rapid-growing industries in the world during the last few decades. The computing power of personal computers has been almost as fast as what you could ever find in high-performance workstations. In recent years, due to the demands for Internet services, computer network technologies have their fastest progresses ever in history. Faster CPUs and better bandwidth has made high performance computing more practical in real-life, especially due to the availability of low-cost computing resources and high-speed computer networks. In this study, an agent-based distributed parallel processing virtual machine is proposed. It is built on PC cluster with popular Linux operating systems. The proposed system is accessible and executable "anytime, anywhere" through standard Web browsers. The agents work on cluster machines that monitor CPU usages and report loads of the corresponding machines to the server. The server dispatches a given job to the appropriate machine according to the performance index and the load of each machine. The job-dispatching algorithm tries to allocate jobs dynamically so that they can be finished in the shortest time possible. The promising experimental results have demonstrated the feasibility and usefulness of the proposed system.