Subject Re: [IBO] Catastrophic Failures on Hyperthreading machines
Author Aage Johansen
Robert martin wrote:
> Hi
> You are right Aage. We have a multi threading App that failed 'only' on
> a HT machine but it actually turned out we had made mistakes configuring
> the components in a multithreaded app.
> However we DEFINITELY do have problems with FB (Super Server) on HT
> machines. The Datapump component takes 24hrs to do a pump that takes
> 10mins on a non HT / Non multi cpu machine.

This is another problem - not a "catastrophic" one. Multi cpu machines
without having affinity for Firebird (SS) set to _one_ cpu are slow - this
is well known. HT may or may not be significant - observations differ
I've just recently switched the HT _off_ on our (2 cpu) database server
(default affinity is _one_ processor, I think), and I haven't seen a
significant change in speed. Admittedly, I haven't tested extensively...
(This is a Dell 2850 with 2 3GHz Xeons)

Aage J.

>>> ...
>>I've been running Fb+IBO on a hyperthreading machine for some time, no
>>problems. Do you try to start multiple threads within _one_
>>connection? I think this will fail, and - with HT - quickly.
>>Aage J.