Re: Help: Reducing Extents -Reply
Posted in 1997
Just a comment about Gregs posting.
I have also used fragment to rewrite a
non-fragmented table in place. I have found
afterwards a high likelihood of receiving 243/172
errors on such tables. This seems to be reduced if
they are [de]fragmented into a different dbspace
that they are currently in. But I have basically
abandoned using it now and gone back to using
clustered indexing which [appears to be | is ] less
flakey.
rgds
Peter Tashkoff <tashkop@zespri.co.nz>
Zespri International Limited
This posting may not be used by any party to vilify
another. Standard disclaimers apply.
>>> Greg <greg@cyberramp.SPAMCHOKER.net>
16/June/1997 04:07pm >>>
Chad,
after increasing the NEXTSIZE in systables (don't
worry too much about the first extent, it'll tend to be
coalesced pretty well unless you have a lot of
fragmented tables in the same dbspace).
two methods not mentioned are:
1. use onunload/onload - it's waaaay faster than
dbexport/dbimport, but takes more disk or tape
space and you can't edit the schema before
onload....
2. use "alter fragment on table <tablename> init in
<dbspacename>" - I really like this one because it
requires only knowledge of the tablename and
where you want it to go. The clustered index
oriented strategies require you to look up stuff for
each table that has little to do with the job at hand.
Alter fragment is much simpler.....
Anyway, you have advice out the keister now, let us
know how it goes....
Greg
p.s. >5 extents isn't necessarily a performance
killer, if you have enough memory to keep your read
hit ratio high (> 99ish percent), you won't be able to
tell the difference between 3 extents and 30
extents. You are at a point it's probably worthwhile
to clean up a little. I try to organize it so I only have
to reorganize things a couple of times a year.
Chad Laity wrote:
snip....
> How do I reduce the number of extents to improve
performance??
> > Thanks,
> Chad Laity
> chad_laity@arrowintl.com
> wick@fast.net
-- Please remove SPAMCHOKER. to get correct
EMail address...thanks.