Mining query logs to optimize index partitioning in parallel web search engines