On Tue, Sep 15, 2009 at 03:39:23PM -0500, Jacob Yocom-Piatt wrote:
> this thread is fucking stupid.
>
> consider that the majority of machines are horribly underutilized, even  
> in large organizations where some of the machines are under heavy load.  
> the reason that everyone here is so dismissive of benchmarks is that  
> they do not translate to real world results. people hyperventilate all  
> day about how software X runs Y% faster under various OSes but i rarely  
> if ever see a concrete expression of this e.g. i switched from openbsd  
> to linux and was able to offer the same level of service with half the  
> machines.
>
> part of the reason that one doesn't normally see concrete examples is  
> that there is far more to the 'performance' of a machine than just  
> benchmarks.
>
> - how does the cost of administration scale with machine count?
OpenBSD sucks at this one. The fact that base isn't packaged is a *huge* pain
if you run lots of it. As is the short support timeline.

> - with what frequency will OS-related issues cause a catastrophic  
> failure in a production environment?
I've never seen any issues caused by the various linuxes we run, nor by OpenBSD
w/r to this.

> - is it easy to upgrade the machines?
Again. OpenBSD really sucks at this one. Building from source is light years
more difficult than 'apt-get update && apt-get upgrade, or 'yum upgrade' or the
like. And you've got to track updates for ports yourself, making those even
more difficult to upgrade.

> - if i don't regularly patch the machines will they get rooted?
>
OpenBSD is at least excellent at this one. But then, it'd want to be given how 
much
of a pain it is to patch. And, it somewhat makes up for it with the fact that
you're going to end up (in lots of cases), running non-base stuff, which will
leave you somewhat vunerable. Maybe less vunerable than $linux, but I don't
think the OS is the worst offender in most situations, by a long way.

> once you start thinking about the answers to these questions you might  
> see how irrelevant most of this discussion has been to date.
>
I still wouldn't say these discussions are irrelavent. If my machines go
faster, I don't have to upgrade my hardware as often, I don't have to use as
many machines, and I don't have to deal with horrible database scaling issues
as quickly. All of these things are useful and important.

> cheers,
> jake
>
>

-- 

-- 

Reply via email to