Re: KAIO vps, and bufwaits

Neil, The number of kio vps is determined by the number of cpu vps configured on your instance. I also think that you get a max queue of 16 pretty quickly. I have seen my queue lengths on a heavily loaded system reach 32. Does anyone know what a high queue length is? Does a highr queue length indicate a disk bottleneck or can adding physical cpu's (and cpu vps) to the box help? Thanks, Doug On Sat, 07 Feb 1998 04:17:27 GMT, jharris@westpac.com.au (Jason Harris) wrote: >Neil, > >Dont forget that BUFFWAITS do not always mean you dont have enough >buiffers. If one thread A wants a record stored in buffer X but thread >B is updating a record that is also in buffer X, thread A gets a >BUFFWAIT. > >Jason > >On Fri, 06 Feb 1998 22:03:42 +0000, Neil Truby <ntruby@netcomuk.co.uk> >wrote: > >>I've recently moved a poorly-performing database server from a Sun Sparc >>20 to a Sun E4002 with 1.5 GBytes of memory, and 2 processors. It's >>running superbly now, but I'm still keen to know where my bottlenecks >>are. >> >>onstat -g ioq consistently shows my kio VP with a maxlength of 16,>>suggesting that the system is I/O bound. In a sample output of onstat >>-g ioq in "Informix Unleashed", three kio VPs are shown, but no >>explanation is given on how to configure a system to use this many (if >>in fact you can). >> >>Does anyone know how to set up extra KAIO vps? >> >>Also, despite having 200M of BUFFERS, onstat -p is still showing some >>bufwaits. The read cache rate is above 99%, but the write remains >>around 87% - the same as on the Sparc when BUFFERS was set to 50M. The >>new system isn't swapping and hardly pages; do the bufwaits suggest a >>further increase in BUFFERS is warranted? >> >>Thanks >> >>Neil Truby >>Chase Manhattan Bank >>Bournemouth, UK >>neil.truby@chase.com