Re: Batch tuning...

Sorry, last message incomplete. I'd agree with most of what William Rice said, but I'd bump BUFFERS up way higher to get your cache rate up; say 200,000. With 4GBYtes of memory you can afford it. Also, look to upgrading to v7.31 or v9.2x. Aside from anything else, you will be able to use the superior kernel asynchronous i/o ("kaio"). regards Neil

In article <932o5s$flv$1@taliesin2.netcom.net.uk>, "Neil Truby" <ntruby@netcomuk.co.uk> wrote: > Sorry, last message incomplete. > > I'd agree with most of what William Rice said, but I'd bump BUFFERS up way > higher to get your cache rate up; say 200,000. With 4GBYtes of memory you > can afford it. > > Also, look to upgrading to v7.31 or v9.2x. Aside from anything else, you > will be able to use the superior kernel asynchronous i/o ("kaio"). > > regards > Neil > > The only reason I am conservative on the BUFFERS is his initial problem was with checkpoint times. From his config it looked like they had BUFFERS=150000 and LRU_MIN/MAX at 3/5 when they were having problems. He might need a new I/O subsystem, or he might need to reorganize his layout, but those can be difficult tasks, one due to cost, the other due to needing to have the time to test and do analyses. Maybe they can get the money back from the consultant and buy a new I/O subsystem. Will Sent via Deja.com http://www.deja.com/

To reduce checkpoint times set LRU_MAX_DIRTY=2 and LRU_MIN_DIRTY=1 with BUFFERS=200,000 this will make the machine do LRU / Background Writes instead of checkpoints. Set your PHYSFILE=2,000,000 which is the maximum so that checkpoints will only be done when the Physical Log is 75% full. Should you wish you can still do your own checkpoints with "onmode -c" or with a Checkpoint Interval of say 3600 (1 hour). With this number of buffers you may be wise to set LRUS=127 and CLEANERS=127 Yours -- Earle A Long (Senior Informix DBA) SINGLEPOINT LTD "William Rice" <ricew@operamail.com> wrote in message news:934noi$iv5$1@nnrp1.deja.com... > In article <932o5s$flv$1@taliesin2.netcom.net.uk>, > "Neil Truby" <ntruby@netcomuk.co.uk> wrote: > > Sorry, last message incomplete. > > > > I'd agree with most of what William Rice said, but I'd bump BUFFERS > up way > > higher to get your cache rate up; say 200,000. With 4GBYtes of > memory you > > can afford it. > > > > Also, look to upgrading to v7.31 or v9.2x. Aside from anything else, > you > > will be able to use the superior kernel asynchronous i/o ("kaio"). > > > > regards > > Neil > > > > > > The only reason I am conservative on the BUFFERS is his initial > problem was with checkpoint times. > > From his config it looked like they had BUFFERS=150000 and LRU_MIN/MAX > at 3/5 when they were having problems. > > He might need a new I/O subsystem, or he might need to reorganize his > layout, but those can be difficult tasks, one due to cost, the other > due to needing to have the time to test and do analyses. > > Maybe they can get the money back from the consultant and buy a new I/O > subsystem. > > Will > > > Sent via Deja.com > http://www.deja.com/ >