home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.databases.informix
- Path: sparky!uunet!cis.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!cbfsb!cbnewsg.cb.att.com!ashaw
- From: ashaw@cbnewsg.cb.att.com (andrew.shaw)
- Subject: I am a stupid (was Online Performance Degradation)
- Message-ID: <1992Sep2.131412.11798@cbfsb.cb.att.com>
- Sender: news@cbfsb.cb.att.com
- Organization: AT&T
- References: <9441@emory.mathcs.emory.edu>
- Date: Wed, 2 Sep 1992 13:14:12 GMT
- Lines: 18
-
- Now that I have had some sleep, I realize that I should expect logN
- behaviour: if the number of adds/minute falls off as log(t), then
- the time for each add increases as log(t), which we are obviously
- never going to better.
-
- Several people have commented that the solution is to drop the indexes,
- do the adds, and then rebuild the indexes. Unfortunately I cannot do
- this, as the adds come in unpredictably, and other processes will be
- doing lookups at all times.
-
- So it appears that Informix is doing a wonderful job, and there is
- nothing else to be done. I think I will try to precisely characterize
- the curve, just to make sure logN is in fact correct.
-
- Thanks to all who responded,
-
- Andrew Shaw
- AT&T Bell Labs
-