--- In firebird-support@yahoogroups.com, Poul Dige <pd@...> wrote:

>
>
>
> > Hi all,
> >
> > our company is developing a system using firebird since years. This system is
> > a "real-time" production steering application.
> > This means for us, that our applications need to perform a lot of different sql
> > (with join over some table) in the shortest possible time. Today if we
> > encounter a slow application behavior, we try to optimize the sql's...
> > I just wonder, what performance would be possible usind firebird, as we are
> > planning to switch to a more flexible database structure, and this will for sure
> > increase the db load.
> > We are using the superClassic with around 30-50 db- users connected. We
> > need to keep our db as small as possible to keep the performance on a high
> > level. This means every year we need to copy away the old data. Actually the
> > fb- process does not often claim to the CPU's limit, so i think the bottleneck
> > are the HDD's, where we are using usuably RAID 1 with fast server HDD's.
> >
> > My question:
> >
> > Is there any complete documentation regarding all the performance critical
> > points for the firebird db?
> > p.e.:
> >
> > What processor to use with which "db- usage- pattern"?
> > -->Up to how many core provide more speed?
> > How to optimally use page size settings (RAM)?
> > What about SSD's? Which modell is good for productive usage?
> > OS- settings, p.e. cluster size of file system?
>
> Our experience with SSD is that the basic speed in case of just a single connection is improved a bit, but once you have many connections the SSD is hardly affected by the number of connections whereas the mechanical disk is suffering very much with the number of simultanious connections.
>
> However, page size and buffer size is quite important for SC/CS-configuration as well - you must definitely test the system with different settings (changing the buffer size is quite easy).
>
> Regards,
> Poul
>

OK, so this means SSD's could be a very interresting alternative to machanical disk, cause we have a lot of users connected.

What about the other points?