Hi !
Have a look at the kernel specs...
you may want to increase the nbuf's parameter size in order to limit
physical I O and to use
more bufferspace.
We have under HP-UX a very good caching siuation alhough we did not
change IB buffer settings at all.

kind rgds
Claus


Thomas Miller wrote:

>Oracle does this automatically. Would it make sense to do this for IB?
>
>Basically, at boot up, Oracle looks to see what memory is unallocated by
>the system
>and then grabs most of it for paging. Mayby and Auto setting instead of
>an exact
>page setting code be used.
>
>Helen Borrie wrote:
>
>>At 12:42 PM 30-09-02 +0200, you wrote:
>>
>>
>>>I wrote:
>>>
>>>
>>>
>>>>I have to work out a -buffers setting that could be a reasonable default
>>>>for both stand-alone installations and multiuser with a server
>>>>(presumably more RAM that the stand-alone thing).
>>>>Given that I have 4KB pages, I think 5000 pages is a suitable value for
>>>>our database, but was wondering: will IB try to allocate upto 20MB (4 x
>>>>5000) even if they're gonna come from the pagefile, or will it stop when
>>>>there's not enough physical memory available?
>>>>I don't want to keep the 75 page server default, but I obviously don't
>>>>want to defeat the cache.
>>>>Does aanyone know for sure? I'm using IB5.6.
>>>>
>>>>
>>Nando,
>>You will defeat the cache if you allocate a default number of pages that
>>exceeds the amount of RAM you have to spare. Assuming the server isn't
>>doing a whole pack of other stuff or running a RAM-intensive admin GUI, the
>>trick seems to be to calculate the maximum amount of RAM that will be used
>>by the maximum number of users likely to be connected doing the maximum
>>amount of work likely to be performed simultaneously. Subtract that value
>>
>>from the physical RAM available, translate that into database pages and use
>
>>that for your default cache.
>>
>>Yes, the cache will spill over into into pag memory if physical memory is
>>inadequate. Since the purpose of the cache is to avoid unnecessary disk
>>i/o it doesn't make a whole lot of sense to set up a cache that will
>>"overflow" constantly to disk.
>>
>>
>>
>>>hasn't anyone got a clue?
>>>
>>>
>>Like so many configuration things, it's likely to be very specific to your
>>users' operating environment.
>>
>>
>>
>>>Or is it a false problem?
>>>
>>>
>>No, just not an easy problem to provide a rule-based solution for. In my
>>experience, it's a thing best handled as part of your soak-testing, where
>>you are presumably well set up to throw a lot of different variables into
>>the mix.
>>
>>It's doubtful you would arrive at a setting which will work equally well
>>for stand-alone and multi-user deployments.
>>
>>heLen
>>
>>
>>
>>To unsubscribe from this group, send an email to:
>>ib-support-unsubscribe@egroups.com
>>
>>
>>
>>Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>>
>>
>>
>>
>>
>

[Non-text portions of this message have been removed]