Subject | Re: [firebird-support] Re: Sort memory |
---|---|
Author | Helen Borrie |
Post date | 2005-06-07T02:35:55Z |
At 11:24 PM 6/06/2005 +0000, you wrote:
page cache for sorting.
With SS, page cache is dynamic, so you aren't going to see the effect of
max. cache unless the cache is "full".
As far as I understand it, sorts are done using dynamic memory from the OS,
if possible, up to the maximum specified in SortMemUpperLimit. However,
that doesn't necessarily mean a sort will be able to acquire that amount of
memory for a particular sort operation. If you have several large order
by's or group by's going on simultaneously and/or the OS is using a lot of
dynamic RAM for something else, someone is likely to cop either an on-disk
sort or RAM being paged out to disk cache during the course of an in-memory
sort.
./hb
>Also, this is for FB SS 1.5.2 on WinXP Pro.The page cache and sort memory are unrelated. The engine doesn't use the
>
>Thanks
>--- In firebird-support@yahoogroups.com, "buppcpp" <buppcpp@y...>
>wrote:
> > When does the sort memory usage come into play?
> >
> > I have SortMemUpperLimit = 67108864 in my configuration, but my
>memory
> > usage never goes above my DefaultDbCachePages settings.
> >
> > I expected the max memory usage to be (DefaultDbCachePages * Page
> > Size) + SortMemUpperLimit when a SORT is involved. But I'm not
>seeing
> > it, so I'm assuming my understanding of SortMemUperLimit is
>incorrect.
page cache for sorting.
With SS, page cache is dynamic, so you aren't going to see the effect of
max. cache unless the cache is "full".
As far as I understand it, sorts are done using dynamic memory from the OS,
if possible, up to the maximum specified in SortMemUpperLimit. However,
that doesn't necessarily mean a sort will be able to acquire that amount of
memory for a particular sort operation. If you have several large order
by's or group by's going on simultaneously and/or the OS is using a lot of
dynamic RAM for something else, someone is likely to cop either an on-disk
sort or RAM being paged out to disk cache during the course of an in-memory
sort.
> >If possible, it's the first choice.
> > So, like I said, When does the sort memory usage come into play?
./hb