An Evaluation of Fault-Tolerant Query Processing for Web Search Engines

A number of strategies to perform parallel query processing in large scale Web search engines have been proposed in recent years. Their design assume that computers never fail. However, in actual data centers supporting Web search engines, individual cluster processors can enter or leave service dynamically due to transient and/or permanent faults. This paper studies the suitability of efficient query processing strategies under a standard setting where processor replication is used to improve query throughput and support fault-tolerance. © 2011 Springer-Verlag.