The document discusses solutions for distributed search on large datasets. It compares centralized search which can be slow for large indexes, to distributed search using Katta, Elastic Search, or an HBase directory. An HBase directory stores the Lucene index across an HBase cluster for scalability and fault tolerance. It discusses indexing and searching phases when using an HBase directory in either lazy or active modes. Performance tests showed the active HBase directory method is suitable but still needs improvement.