Web-Based Performance Monitor for Distributed Computing

Abstract

There are lots of distributed computing styles, such as Cluster, Grid, Cloud Computing and so on, based on the larger scale network popular Infrastructure. It is becoming more and more complex to common users in terms of both massive nodes resources management and parallel applications optimization. Performance monitoring tools have received a great significant attention. On the one side, it is necessary to support managing increasingly complex hardware and software infrastructures; on the other side, it can provide user system load and application running information in real time so as to guide user's resource request scheme and program optimization. The hot research issues are associated with some important software properties, such as scalability, elasticity, adaptability and timeliness. Based on the ADS principle, to meet large-scale and continuous monitoring of applications online, this paper proposes a Web-based scalable performance monitoring system framework with a flexible configurable network to address the massive amounts performance data obtained in timeliness. An online transport network is constructed in charge of collecting and processing performance data during execution from different node. It can make the global information available at runtime. Several initialized strategies of Tree-based overlay network are discussed in different cases. A tool has been developed and experimented, and it has been to be effective by many statistics.