Subject Re: Closing the gap between OAT and Next transaction
Author alex_vnru
--- In ib-support@y..., "Jason Wharton" <jwharton@i...> wrote:
> > No, this gap always present if you have connected application
with
> > opened select. Any statement is performed within transaction
context,
> > some connectivity components mask it.
>
> This is true of component sets which do not have an architecture to
use
> server resources efficiently.
>
> This isn't true of IBO.

Jason, I had'nt in mind to criticize IBO, just pointed that in C/S
systems client connectivity software is subsystem with own behavior
which can be realized in many ways, not a thin streak. When I said
this I had in mind BDE.

> I wouldn't tolerate a gap of 30,000. If you knew what overhead this
imposed
> upon your server you wouldn't do it any longer.

I know my design is'nt the best in this case. It is large
applications migrated from BDE and decision to hold long read commited
read only transactions was compromise between perfomance and migration
time. There are near to nothing delete operations and update on
indexed columns ones in this applications. And I assure you - we don't
feel any performance degradation during a day. More, we don't feel
difference since RC2 was installed and gap stabilized at level about
500 with peak value about 2000. We did'nt made exact tests, I speak
about subjective impression.

Best regards, Alexander V.Nevsky.