DB2 Version 10.1 for Linux, UNIX, and Windows
DB2 Version 10.1 for Linux, UNIX, and Windows
A number of factors influence performance and resource use in DB2® Text Search. When planning system capacity for DB2 Text Search, consider the query workload, the number of parallel index updates, the expected size and growth rates of your text indexes, and the processing time for the documents you are indexing.
DB2 Text Search enables full-text search queries on most data types within the DB2 database, including support for XML documents and a rich-text or proprietary format feature. Full-text search is supported through a text search server instance that is integrated with the database instance or in a stand-alone setup associated with the database instance. Communication between the database and text search server instance is through TCP/IP. Full-text indexing and search performance depend on the text search server configuration, available system resources, and text index specific settings.
A single text search server is configured for the database instance. The text search server has a recommended minimum memory requirement of 4 GB of memory for production use, which increases according to the number of parallel index updates.
Updating the text search index is resource-intensive, both in terms of disk I/O and CPU or memory requirements. Multiple configuration parameters are available to control the Text Search server resource usage. For workload distribution, for example, in a partitioned database environment, a stand-alone setup is recommended.
On average, a text search index is about 50-150% of the original data.
There is no absolute size limit for text search indexes, however, the combination of throughput factors with completion time dependencies results in practical limits on the total text search index size. For example, when a considerable amount of data is added to or removed from a text search index, the text search index structure is merged to improve query performance, and the time for completion of the merge depends on the size of the index.
Absolute text index update throughput depends on the data type and the index format. For perceived query performance, the biggest impact is due to the number of matching results, not the size of the text search index. For example, a query with a single predicate using a single-term search term on a 100 GB text search index performs similar to a search on an 800 GB text search index if the number of results is the same.
Optimal processing for text index updates occurs when there is approximately 10-100 KB of text per document. Throughput degrades above 1 MB and below 1 KB of text.