Mining Query Logs to Optimize Index Partitioning in Parallel Web Search Engines