A service-oriented monitoring system with a forecasting algorithm of a time sequence-based hybrid model1

Abstract

In dynamic and autonomic grid environment, monitoring of heterogeneous resources is critical for many grid functionalities, such as performance analysis and tuning, job scheduling, fault detection and diagnosis. Although, there is much literature on grid monitoring research, the inter-operations between different monitoring systems have been always neglected. Besides, many monitoring systems introduced too much intrusive overhead. GridEye is a novel service-oriented monitoring system with a flexible and scalable service-oriented architecture. An open and extendable information schema is adopted to standardise data contents; Web Service interfaces are employed to provide inter-operative ability for different grid systems. As a feature, a time-sequence-based forecasting algorithm is provided for precise performance prediction. Experiments showed that the performance is comparable with other famous monitoring systems. We also proved that GridEye introduces little overhead and achieves accurate forecasting.