Re: (2) Poor Performance with 7.31.uc6

Obnoxio The Clown wrote: > > From: "Mark D. Stock" <mdstock@mydas.freeserve.co.uk> > > > >Obnoxio The Clown wrote: > > > > > > A word to the wise: when you need to do your index rebuilds, set the > > > following environment variables: > > > > > > PSORT_NPROCS=[2 * NUMCPUVPS] > > > PDQPRIORITY=100 > > > PSORT_DBTEMP=/tmp1:/tmp2:/tmp3 # i.e., 3 temporary filesystems > > > > > > You will be mildly impressed by the index build speed. I've been fooling > > > about with a 100M row table, and it builds a single-column index in < 20 > > > minutes. On a 2-way HP box. > > > >Don't limit yourself to 2x. It really depends on CPU speed and disk > >speed / number. Also fragment the table and or index if you can. > > Some of us are working on weenie HP boxes, remember. :-) I feel sorry for you. ;-) > I think fragmenting the table is more important than fragmenting the index. > But that's just me. :) Yes it is. Also, don't be shy to place several fragments on the same spindle. Disks these days can handle far more throughput than most people give them credit for. The ultimate aim is to get IO keeping up with CPU. I have 4 CPUs stuck on 99.90% during an index build, and for this build we don't care how long it takes to type 'ls'. :-) Cheers, -- Mark. +----------------------------------------------------------+-----------+ | Mark D. Stock mailto:mdstock@mydas.freeserve.co.uk |//////// /| | http://www.informix.com http://www.informixhandbook.com |///// / //| | http://www.iiug.org +-----------------------------------+//// / ///| | |This email will self-destruct in |/// / ////| | |10 sec. If you received this email |// / /////| | |in error, sorry about the mess. |/ ////////| +----------------------+-----------------------------------+-----------+