Subject Re: [firebird-support] Re: firebird classic on windowx 2000|xp|2003
Author Helen Borrie
At 09:35 AM 28/01/2008, you wrote:
>Hi Hellen,
>
>> Could you please go to the very first page of the release notes
>> that you are looking at, and report the document release number
>> and date, please? Classic was experimental at the beginning
>> (Fb 1.5.0), but that was almost 4 years ago. So let's see
>> whether we need to correct this warning for some or all of
>> the notes for the five sub-releases since then...
>
>I went to the release notes of firebird 1.55 (the latest) and
>"experimental" is yet there.

Thanks - will fix.

>About page cache (DefaultDbCachePages) with page size=4k, I well
>understood for evident reasons that high value is impossible on classic.
>However, on SuperServer, You precise that this conf entry must be let
>between 128 and 512.

No! we were not talking about Superserver but about Classic! For superserver, it is reasonable to set the page cache to e.g. 8000 pages, depending on the page size of course. 8000 pages would be out of field if your page size were 16K.

>Even when the server has 4Go ram (1 mega pages),
>does this rule must be observed , especially if there is no other
>application, as I mentionned?

It seems you need to study the messages more closely and also do some arithmetic! Multiply no. of cache pages by page size and you will arrive at the size of the cache in BYTES. It makes no sense to set the size of the cache so high (on either model of Fb) that the cache is being constantly swapped out to disk by the OS memory manager.

>It would be obvious that the caching pages is done in ram since memory is yet highly availabe. I don't see
>any reason making the os playing that on disk while lot of ram is free.

Don't overlook the fact that the OS will not make more than 2GB of RAM available to *any* single 32-bit process.

>So let me please ask again my question : why DefaultDbCachePages=20480
>on FbSS (with well dimensionned server entirely devoted to firebird)
>could by a "typo" ? In other words, if there is no risk to occur on
>disk but on ram, is it a good practice to put so high value ?
>thanks.

I hope you have figured it out by now. :-)

./heLen