Developing Performance Monitoring and Analysis Middleware Based on the Network Weather Service

University Of California-Santa Barbara, Santa Barbara CA

Investigators

Abstract

As the Internet evolves, so must the monitoring capabilities. This is especially important for evolving applications which require dynamic resource allocation such as collaborative applications, GRID technologies and distributed computing. In addition, because of the complexity of the network and the rapidly changing requirements of the applications, it is very difficult to predict performance or capability. This proposal plans to develop a scalable performance monitoring and dynamic prediction infrastructure based on The Network Weather Service (NWS). The NWS operates a distributed set of performance sensors, applies a set of fast statistical forecasting techniques in near real-time, and generates forecast reports for the resources being monitored. The implementation is portable among all the Grid middleware infrastructures. The proposal has two research foci; 1. research and development to extend NWS Scalability 2. deployment as research infrastructure The goal will be to develop and deploy The Network Performance Oracle which will provide a persistent, internet-accessible service for forecasting end-to-end network performance.