[ILUG] Favourite flavours

[Bryan O'Donoghue]

Brian Scanlan wrote:
> On Mon, 12 Jul 2004 in lists.ilug you wrote:
> 
>>Now while I broady accept the argument that for hosting companies like 
>>blacknight, it is essential to be able to put up a production box in <= no 
>>time... the world is not a vax.. or rather a hosting company.
> 
> 
> It sounded like he wanted a pair of servers for a production environment.  From
> what I understand, Gentoo is fun and can be used in production if you're quite
> comfortable with it and linux in general. I'm not sure if I'd reccomend it to a
> newbie though - Debian and Fedora spring to mind first.


Debian is hardly noobie friendly, also, wasn't Linus supposed to have been C 
programming for 1 1/2 years when he wrote the first Linux Kernel?

/tounge cheek

>>I have nothing against Red Hat boxes 
>>really, but, like I say, if you have the time and patience, you'll be hard 
>>pushed not to see better performance from a arch specific optimised box.
> 
> 
> What would you estimate the performance gain to be, and how long will the
> servers be spending compiling each package (and updates to packages) to get
> this gain? 

I've never benchmarked it, and I accept that it is therefore possible (but, 
extremely unlikely) that a glibc/kernel/java runtime/httpserver compiled for an 
i586 would be as fast or faster, then say such a system compiled specifically 
for a P4 for arguments sake. Possible, but, I hope you'd agree, probably unlikely.

However, it's probably worth noting that, if there was no performance gain to 
be had, distros like Gentoo, probably wouldn't exist. Maybe they would, 
probably they wouldn't.


>*And* how much of this performance gain is relevant to the task at
> hand (i.e. a java application server, presumably running a pre-compiled jvm). 

Even still, if you aren't using something like blackdown (can that be used in 
this context, I don't know), you'd still *probably* be able to get a gain... in 
performance, how much of a gain is debateable.

However, you've missed the second point in Gentoo's favour.
Portage.

Portage is *excellent*, at least, I like it.
I don't know art, but, I know what I like.
*grin*.

>I
> don't think the "low hanging fruit" in this environment will be a lack of
> processor specific binaries *should* a performance issue arise.
> 
> The only gentoo benchmark document I can find is this...
> http://www.gentoo.org/main/en/performance.xml
> ...you'll find that most production servers don't spend their time in "load
> time", 

That *seriously* depends on the Server and the applications it's running.

A DB server for example, on a busy website can spend quite a bit of time in a 
high load situation. That's why you get it ram... and maybe even end up doing 
things like tweaking it's kernel parameters, to allow higher levels of 
semaphore, more shared memory... the ability to make coffee in your colocation 
facility et-al.

> and why they didn't compare a gentoo install with no optimisation to one
> with arch specific optimisation? Compiling for a specific pentium
> architecture/revision is hardly ever worth the time, unless you're running
> specialised applications (i.e. not basic web serving and java application serving).
> 
> Brian.

Err, didn't we just decide that having an optimised jvm was highly desireable?

Somehow, I'm going to have difficutly believing a sizeable server application 
will not run any quicker compiled with -O3 -march=pentium4, then it would with 
-mcpu=i586.

-- 
Bryan O'Donoghue
Embedded Software Engineer

Europlex Technologies Ltd
Clonshaugh Business & Technology Park
Dublin 17
Ireland

T:+353 (0) 1 2500500
F:+353 (0) 1 2500590
E:bryano at euoplex.ie
W:www.europlex.ie