[Total OT] Trying to improve some numbers ...

lars lars at gmx.at
Fri Feb 17 10:52:41 PST 2006

Marc G. Fournier wrote:
>>> You can say you are losing out on 'stability fixes', else the server 
>>> itself wouldn't stay up that long ... so about the only thing you 
>>> lose would be performance related improvements and/or stuff like 
>>> memory leakage ...
>>> And I could do this all *without* any firewalls protecting it ...
>> Even if you managed to maintain an old version of a particular OS's 
>> uptime for so long, what did you prove?
> Wasn't arguing that I "proved" anything, only that a long uptime could 
> be achieved *without* any security implications :)
Point taken :-)

>> IMHO 'uptime' as a 'feature' is overrated, not to say obsolete.
> Agreed 100% ... Availability is the useful metric, not how long a 
> stretch of time the OS can remain running ... not necessarily worded the 
> best way, but our uptime policy (http://www.hub.org/uptime_policy.php) 
> was such that we tried to upgrade our servers once every 30 days or so 
> ... not always possible, and lately less so, but it was our aim ...
Actually it sounds quite reasonable.

I used to work for a major IT corporation and their SLA didn't
amount to much more than that in that particular class of service
(i.e. not highly available and/or clustered machines).
But they needed a lot more words, to spread the wealth to the legal 
departments of all parties involved.

I'm not entirely against such efforts of long uptimes.
I strongly believe in efforts to back up rumor with fact, as in
the rumor 'FreeBSD is rock-solid'.
Actually I believe it is, but I can't prove it beyond talking
of my own experience with it.

IMHO a lot of so-called fact is actually hear-say or anecdotal,
either because the people spreading these 'facts' don't bother to
update their info, the info is purely from each sysadmins personal
experience and/or because there's a lack of info that is standardised
and repeatable.

So this web-site's effort, although similar to the Netcraft uptime stats, 
is quite alright since it's a first step towards getting some numbers.

More information about the freebsd-questions mailing list