162 Siebel 7.8 with IBM DB2 UDB V8.2 Handbook
Increasingly, implementations have large tables, users covering numerous time
zones, and must keep maintenance windows to a minimum. Determining what
maintenance is necessary and how frequently it is necessary will maximize the
usefulness of available downtime.
A simple approach is to track the larger and more active tables in an
implementation. For example, the sample script in Example 6-17 on page 158
can be modified to check for cluster ratios or normalized cluster factor of 95%,
90%, and 85%. For example after running RUNSTATS, we see that S_CONTACT is
95% clustered on its cluster key, S_CONTACT_U1. Two weeks later,
S_CONTACT_U1 crosses the 90% clustering threshold. Database administrators ...