Three-level caching for efficient query processing in large Web search engines