Index Compression with DB2 9 for z/OS
An IBM Redpaper publication
Note: This is publication is now archived. For reference only.
In data warehouse applications, it is common to have many very large indexes defined on large tables. It is also possible that the index spaces exceed the size of the tables upon which the indexes are based. As one way of helping you to manage the cost of ownership, in
DB2 9 for z/OS (New Function Mode) you can choose to compress your indexes and you can define indexes with a page size larger than
4 KB. All references to the DB2 product refer specifically to DB2 9, except where the release level is explicitly mentioned.
Compression has been available for table spaces on z/OS since DB2 Version 3. This IBM Redpaper explains some of the significant differences between data and indexes in the way that compression is managed, as well as a comparison of the performance characteristics.
A limited set of performance measurements obtained in a laboratory setting are described in this paper, but the main spirit of this paper is not about specific performance numbers. In other words, the paper describes how the implementation methodology affects the performance of both data and index compression, and the authors hope that the reader does not get too focused on the path lengths of the implementation or the performance of any specific hardware. User experiences will vary. It is difficult to predict the impact of index compression on performance, but in this paper we attempt to provide a methodology that will give some insight about how to predict the effect of index compression on the CPU utilization.