Study on Grid Resource Monitoring and Prediction

摘要

Grid is a collection of heterogeneous systems which share the computing power and storage capacity over the internet. Grid applications usually involve large amounts of data and/or computing and often require secure resource sharing across organizational boundaries. Thus they cannot be easily handled by todays Internet and Web infrastructures. Monitoring the resources in a grid system is necessary to determine any performance problems and for tuning the system for the purpose of optimizing its performance. Prediction data about the resources and applications are needed for managing the various grid resources more effectively. This will also help in scheduling the jobs more efficiently. Various prediction and monitoring systems are available which perform either prediction or monitoring or both for grid systems. Since the grid systems they aim at are different, they have very different characteristics, architectures, and use different techniques. Predicting the dynamic availability of resources such as CPU time, network bandwidth, and disk bandwidth is vital to load-balancing. In this paper, a comparison of various prediction and monitoring systems for grid environment is done.