Dave Page wrote:
> Alban Hertroys wrote:
>> Joshua D. Drake wrote:
>>>> I agree with you on the multi-threaded. I think I will add a note
>>>> saying the the multi-threaded architecture is only advantageous on
>>>> Windows.
>>> And Solaris.
>>
>> I'm not entirely sure what makes multi-threading be advantageous on a
>> specific operating system, but I think FreeBSD should be added to that
>> list as well... They've been bench marking their threading support using
>> multi-threading in MySQL (not for the db, mind you - just for load ;),
>> and it performs really well.
>>
>
> I'm not sure I necessarily agree with those two - we have no real proof
> that a multithreaded architecture would be significantly more efficient
> than a multi process. It certainly wouldn't be as robust as an error in
> one backend thread could bring down the entire server.
>
> Windows is a special case in this regard. The OS has been designed from
> the outset as a threaded environment. The important point is not that
> Windows threads are necessarily any more efficient than their Solaris or
> FreeBSD counterparts, but that the multi-process architecture is alien
> to Windows and is inherently slower. Two of the major bottlenecks we
> have on Windows as a result are backend startup time and shared memory
> access speed - both of which are significantly slower than on *nix.
>
> Regards, Dave
Thanks for explaining (again).
So actually the remark shouldn't be that "the multi-threaded
architecture is only advantageous on Windows", but more like "the
multi-process architecture is disadvantageous on Windows and hence a
multi-threaded architecture is preferred (on that particular OS)".
--
Alban Hertroys
alban@magproductions.nl
magproductions b.v.
T: ++31(0)534346874
F: ++31(0)534346876
M:
I: www.magproductions.nl
A: Postbus 416
7500 AK Enschede
// Integrate Your World //