Re: Do any of you have a policy on rebuilding indexes?

Patrik:

Disk fragmentation does not exist in IBM i due to single level storage, so
that's not an issue. If there is an imbalance of disk utilization IBMi
will over time balance that out all on its own. That's not to say that
files themselves can't be degraded due to deleted records. Most access
paths are set to immediate update, so if there are ones that are not in
use, it helps to get rid of them to lessen overhead, but again on current
systems I doubt that will have much of an effect on performance.

If disk units were added to an existing system we would run STRASPBAL for
capacity and the system would then move things around based on its own
understanding. If you really want to balance based on usage, then you need
to run TRCASPBAL first so the system can collect the information it needs
to balance that way. I've not done that in many years since the storage is
just too fast to notice any degradation based on usage. With the new use
of NvME and SAN storage I have not done either of these things in quite
some time.

On your system even though it is older you would be able to see these
things in action if you are able to add a drive.

--
Jim Oberholtzer
Chief Technical Architect
Agile Technology Architects


On Thu, Mar 30, 2023 at 10:12 AM Patrik Schindler <poc@xxxxxxxxxx> wrote:

Hello Mark,

Am 30.03.2023 um 14:48 schrieb Mark Murphy <jmarkmurphy@xxxxxxxxx>:

We have a client who wants to rebuild indexes, just because they haven't
been built for a long time (3-4 years). They think it will improve
performance. I've never rebuilt indexes to improve performance. Does

anyone

have an opinion on this?

In theory, any object might become fragmented on disk over time.
Recreating indexes thus might increase performance.

If this theoretical increase is relevant in practice, or even measurable,
I can't tell. This heavily depends on the hardware details such as disk
count (and type: mechanical or SSD) of the affected ASPs, amount of RAM,
etc., and on the size of the tables and indices themselves.
Probably there's also a penalty for writes, eventhough there's enough RAM
to keep the index there (for reads). This also depends on usage of the
tables in question, and if journalling is in use (which AFAIK buffers table
writes in RAM).

Not an easy to answer question, as you can see. Being a while in this
group, I can't recall coming across the topic of defragmentation on IBM i.

I disagree with Rob. I can't see how this might end up being a shoot in
the foot.

:wq! PoC

--
This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list
To post a message email: MIDRANGE-L@xxxxxxxxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: https://lists.midrange.com/mailman/listinfo/midrange-l
or email: MIDRANGE-L-request@xxxxxxxxxxxxxxxxxx
Before posting, please take a moment to review the archives
at https://archive.midrange.com/midrange-l.

Please contact support@xxxxxxxxxxxxxxxxxxxx for any subscription related
questions.