Enabling Queries Using the Grid-brick Approach: A Distributed Data Storage Architecture

This paper presents a Grid-based parallel Event Processing System (GEPS) that is applicable to all the domains where large collections of separate blocks of data (images) need to be processed, stored and analyzed. Data intensive applications are becoming increasingly important in many science areas. In many domains the need for future computing and data management capabilities is difficult to accommodate within the expected technology improvements and the usage of scalable solutions, distributed over a large number of nodes, becomes essential. The processing, storage and analysis of High Energy Physics collisions illustrates the main challenges of this research area. Using the Globus grid toolkit, we have developed the GEPS system that provides a framework for data storage, processing and analysis that is based on the Grid-brick approach. The main concept of this approach is that individual farm nodes mirror the same architecture to provide all the available services for the data stored locally that is to each node. The data is split over different nodes that are responsible fro processing the local queries. Performance results indicate that event processing and filtering can be effectively implemented on GEPS, encouraging the continued effort to improve the GEPS prototype.