Metadata repository sizing

The information contained here is based on tests with a Db2 database. IBM® InfoSphere® Metadata Asset Manager imports metadata assets from external tools, files, and databases into the metadata repository of IBM InfoSphere Information Server. The metadata is stored in the repository in the USERSPACE1 and XMETA tablespaces. As a result, the disk space allocated for USERSPACE1 and XMETA grows over time with additional imports. The metadata repository is used to store other information as well, such as project, report, and data store connection information.

The metadata repository is a schema in the metadata repository database. The following information can be used to determine the sizing requirements for the metadata repository alone. If the metadata repository database contains schemas for other data stores in InfoSphere Information Server, refer to capacity planning topics for those data stores to determine the total sizing requirements for the metadata repository database.

The initial size of the Db2 metadata repository database is approximately 1 GB. The following calculations can be used to determine how much data will be contained in the metadata repository schema. As that amount grows beyond 1 GB, the Db2 database will need to expand to accommodate the data.

InfoSphere Metadata Asset Manager assets

Experimental results based on a broad range of imports against various workloads (using IBM Netezza, IBM Db2. CA ERwin, IBM InfoSphere Data Architect, IBM Cognos, and SAP BusinessObjects) indicate the following behavior:

  • For each 1,000 assets that you import by using an express import, the amount of disk space used by XMETA and USERSPACE1 grows by approximately 5 MB.
  • USERSPACE1 typically uses 65% of the total space, while XMETA typically uses 35%.
  • Each reimport increases the disk space used by approximately 55% of the disk space change from the initial import. (For example, if the initial import uses 10 MB, a reimport would typically use an additional 5.5 MB.)

This behavior can be used to estimate disk space used by the metadata repository. The following example estimates the effect of importing 10,000 assets once, and then reimporting them 10 times:

Initial import: (5 MB per 1,000 assets) * 10,000 assets = 50 MB
Reimport: 55% * 50 MB * 10 reimports = 275 MB
Total disk space used by XMETA and USERSPACE1 = 50 MB + 275 MB = 325 MB

Of that 325 MB, 65% will be used by USERSPACE1 and 35% used by XMETA.

IBM InfoSphere DataStage® assets

Experimental results indicate that 1,000 DataStage assets (such as jobs, project, containers, and routines) require about 250 MB of disk space.