實作一個在UNIX環境上的資源監控系統

Abstract

在UNIX環境中，同時使用多台的工作站，若不能有效的分配負載，不僅造成工作站使用率低，還會導致服務品質不佳。為了提高使用率，許多使用者會採用Load Sharing系統來改善。然而，市面上的Load Sharing系統皆只針對CPU的utilization來做管理，使用對象侷限在designer的batch job，對其它工作站資源（例如：Memory）及使用者的便利則考慮不多。 在本篇論文中，我們提供了一套資源監控系統給所有使用者。它可以掌握所有工作站資源的使用狀態，讓使用者輕鬆地挑選可利用的工作站來作業，減少因多人共用一台工作站導致其它工作站閒置的情況發生。這套資源監控系統以perl/Tk來撰寫，架構為client-server，它提供直覺且友善的GUI介面給使用者。同時它採用socket方式來傳遞資訊，並每60秒更新資訊一次。除了上述功能的提供外，我們也根據軟體的使用性、容錯能力、效能及重複使用性分析，確保它的品質。

In an UNIX environment, if the loads cannot be distriputed effectively, the execution in multiple workstations might have several defects, e.g., low usage rate in workstations, poor service quality, etc. Currently, many users apply load sharing system to enhance the usage rate. However, commercial systems of resource sharing managements are based on CPU utilization and their functions work on the management of designer’s batch jobs only. To user, the usability and resource accessing of other workstations are less concerned. In this thesis, we describe a resource monitoring system, RMS, to solve above problems. RMS is designed to monitor the usage of all the workstation resources. By providing an intuitive and friendly interface, RMS can help reduce the overall loading in the whole system since the available workstation(s) can be easily selected to work. The second part in RMS is to update related information in each workstation (e.g., CPU status, memory) every 60 seconds. This part can clear the garbage process/memory and help increase utilization too. RMS is developed with perl/Tk and based on client-server architecture, where socket method is used for data exchage. Finally, we also analyze usability, fault tolerance, efficiency and reusability to ensure its quality.