Information Pooling And Archiving Mechanism In Grid Monitoring System

Grid monitoring system aims at collecting real time information of resources and platforms, in order to support the performance of grid. There are many problems in present grid monitoring systems, such as the information converged is not comprehensive, the methods of aggregation are too simple, the relative analyses and archive modules can not completely meet users'demands. Consequently, the research on Information Pooling and Archive are very valuable.Data pooling and Data Archiving are two facets in grid monitoring. Monitoring data pooling in real time demands clients to be able to retrieve and display complicated information of distributed monitored clusters and implement the Subscription/Notification(S/N) mechanism with high efficiency. This design integrates not only the general displaying-data-in-charts function, but also GIS (Graphical Information System) technology. It is implemented with the goal of providing S/N mechanism as smoothly as possible. The number of its monitoring metrics is up to 50. This system also embeds the functions of monitoring grid platforms, such as monitoring grid services, grid users, grid jobs, grid states, etc. In addition to display monitoring data on clients'end in visual effects, the system also stores all data in an archive module. The goal of Grid Monitoring Archive Mechanism is to provide users with high qualities, including classifying and compressing large-scale historical monitoring data, and reduce the load of network. The purpose of classifying data is to calculate appropriate ratios for the time series data to be compressed. We adopt a sampling compression method to compress monitoring data and design several brand-new index algorithms (including Peak Focus Reduction, Adaptable Interzone Sampling Reduction, etc) with high effect and high efficiency.We present high quality and efficiency of our information pooling technology used in the monitoring system, including converging data in real time, integrating monitoring system with GIS, dynamically monitoring clusters and network, monitoring grid platforms, etc. With performance testing, we present the high qualities and efficiencies of our compressing algorithms. The compression ratio with our method is able to be up to 95% and keep the distinct fluctuations to the corresponding original time series. Consequently, the results could prove the high quality and efficiency of this kind of archive mechanism.