Performance evaluation of a distributed architecture for information retrieval

Information explosion across the Internet and elsewhere of-fers access to an increasing number of document collections. In order for users to eectively access these collections, in-formation retrieval (IR) systems must provide coordinated, concurrent, and distributed access. In this paper, we de-scribe a fully functional distributed IR system based on the Inquery unied IR system. To rene this prototype, we implement a exible simulation model that analyzes per-formance issues given a wide variety of system parameters and congurations. We present a series of experiments that measure response time, system utilization, and identify bot-tlenecks. We vary numerous system parameters, such as the number of users, text collections, terms per query, and workload to generalize our results for other distributed IR systems. Based on our initial results, we recommend simple changes to the prototype and evaluate the changes using the simulator. Because of the signicant resource demands of in-formation retrieval, it is not dicult to generate workloads that overwhelm system resources regardless of the architec-ture. However under some realistic workloads, we demon-strate system organizations for which response time grace-fully degrades as the workload increases and performance scales with the number of processors. This scalable architec-ture includes a surprisingly small number of brokers through which a large number of clients and servers communicate.