I have a client using DBMS V6.1-2 on VMS 6.2 with a curious record
storage problem. In the middle of the day, the number of pages checked
to records stored ratio goes to about 370 to 1. Normally it is about
1.2 to 1. On this node of the two VAX cluster, a single application is
running and storing only one type of record to a single area. The
problem goes away when most of the users log off the other VAX, which
may be coincidental.

Normally, an excessive number of pages checked for each record stored
indicates the Space Area Management (SPAM) Thresholds are not set
properly. I think they are set to appropriate values, and have been
working well for years. What has me stumped is how the behavior
changes.

If you are familiar with MANMAN, the application is storing material
transaction records (MATREC). The storage placement is SYSTEM DEFAULT,
area size 300,000 pages, page size 12 blocks, SPAM interval 24,488 (13
SPAM pages). All three thresholds are set to 79%. The static record
size is 624 bytes and the record has no dynamic elements. The area
does not have snapshots.

Any ideas would be much appreciated.