Subject Re: [IB-Architect] SUPERSERVER_V2
Author Dmitry Kuzmenko
Hello, Mark!

Mark O'Donohue wrote:

> I'd be interested as well in seeing the results, and also I think

I'll prepare results after second test, so there will be things to be

> interested in the test scripts, one thing we are missing is general
> load tests, and this might prove a good place to start.

sorry, no scripts. I've just copied CreateFile call from sources,
and made simple sequential and random read/write buffered operations.
I.e. simulated working with pages without internal cache.
The tests were run only to understand how different CreateFile flags
(page size and cluster size also) affects read/write performance.

> On another note, although not suitable for end users I think it might
> not be a bad idea to have some of these things available via switches to
> developers in a debug build.

I've found Jim's rule very good, and now I see that no_buffering is unusable for
IDE dives. Maybe after some another tests (on another IDE and some SCSI)
I will change my mind, don't know.

> Real programmers like to live on the edge ;-).

Of course! :-) Another story is how Interbase will work with these
"special found flags". Plain tests I've made shows only HDD performance.
The only why I've asked to insert no_buffering is because I can't compile
sources yet by myself. Sorry, I'm not C/C++ programmer.
But, seems that it is easier to install MSVC than to wait for somebody who will
add new untested (!) functionality and prepare new distributive for public (!)

Dmitry Kuzmenko, Epsylon Technologies.