Avoiding long outages with large databases

Discussion created by ca.portal.admin on Aug 8, 2007
Latest reply on Aug 8, 2007 by ca.portal.admin
I have a question regarding large databases. In the last few years for
various business reasons, we have drastically increased the number of
occurrences of 3 specific record types (two are via the third) that
reside in the same area. The numbers will continue to increase. Our
DBA has increased the size of the area several times to accommodate
this. However, she has advised us that the area cannot be made much
larger, as it will mean long outages when database maintenance/changes
need to be done involving that area (restructures, etc.). Our business
runs 24/7, and so long outages are a major issue for our users. One of
the records in question is 464 characters, and we have almost 2 million
occurrences of it. It is used frequently by our applications. The
second is 168 characters, and we have 1.5 million occurrences, and the
third is 52 characters, and we have 1.5 million occurrences. The area
currently has over 1 million pages. We do archive/delete old data, but
not enough is considered 'old' by our users' requirements to alleviate
enough of the problem. No doubt there are larger databases out there
than ours with more data...what strategies/products/etc. do you use to
avoid long outages when you must make changes to the records/indexes in
an area? Any feedback would be most appreciated.

We are a VM/VSE shop with IDMS Release 15.0.

Confidentiality Notice:
This Informatix Incorporated email message is intended only for the
use of the individual or entity to which it is addressed, and may
contain information that is privileged, confidential, and exempt
from disclosure under applicable law. Access to this email by
anyone other than the intended addressee(s)is prohibited
without prior approval.
IDMS Public Discussion Forum


Re: Avoiding long outages with large databases
"My understanding is that it is more costly in terms of database I/O if the 'via' records are in different areas than the record they are related to. Could there be performance issues associated with splitting up the area that way?
Linda Campbell <Linda.Campbell@INFORMATIXINC.COM> 8/8/2007 12:40 PM >>>
Been a while since I worked on VM, but when I did there was 4K page size
limit and that unfortunately increases the number of pages and can
impede effective clustering. If that is still true (or even if it's
not), one possible solution is splitting the single area into 3 separate
areas, each containing one of the record types mentioned. By doing
this, you could improve throughput by spreading out I/O access and
reduce the number of pages assigned to each of the individual areas.
You might also get away with doing maintenance on only one of these
areas at a time depending on the growth patterns of the records. This
can cause some significant application changes because you would need to
add the extra areas to any programs that use these records, but it might
be worth it if this area will continue to grow at a rapid pace.

Linda Campbell
Informatix Inc