An Investigation into Query Throughput and Load Balance Using Grid IR

Abstract

Grid information retrieval (GIR) means using a grid system for retrieving relevant documents that satisfy the user need from within a large-scale data collection. A data collection can be text, audio, video, etc. The grid provides powerful computation while information retrieval provides techniques for retrieving useful information. In previous work, we built our baseline system and we compared three types of IR systems, namely document-based, term-based and hybrid partitioning with respect to average query response time. In addition, we proposed new methods for improving the load balance and query throughput for term-based and hybrid partitioning. We carried out a set of real experiments using a broker and six nodes. In this paper, we propose to repeat our previous experiments using the grid as a new IR environment.