I've a feeling this could go on for some time, so I'll stop it now (at
least the posts from my side).
Our approach works for us, and our users receive the service we intend
for them. Yes, we could maintain a greater uptime if it were required.
I'd be very surprised if there were any noticeable degradation even if
the server remained up for a month (arbitrary figure), but we have the
luxury of not having to wait to find out if in fact server performance
would in fact dip :)
T
-----Original Message-----
From: Adrian Midgley [mailto:[log in to unmask]]
Sent: 09 December 2004 13:59
To: [log in to unmask]
Subject: Re: rapid administrative response to unreliable NHS website
On Thursday 09 December 2004 08:11, Terry Brown wrote:
> "A machine that is restarted every two days is a machine whose owner
> doesn't believe can run for 4 days." [Midgley, 2004]
> I've stated this once already, but it appears to have been ignored.
I noted it, and continue to disagree. (Not specifically about the
margin of
safety or max duration or whatever)
If you were not sure it would run for 2 days, you'd reboot it after 1
day.
> I'm
> 100% sure that I could maintain a significant uptime on the servers -
I
> just personally see no need to introduce any risk, whatever the
> operating system, to optimal performance from a system.
Does not compute.
> This was initially started as a 'bash microsoft' point, and to be
honest
> I'm keen for those to exist wherever possible, but our choice of
> rebooting alternately each system was not born out of some latent fear
> of the Microsoft platform, but of wanting to maximise the performance
of
> the servers that we are delivering a service from, and having enough
> hardware that availability, and not uptime, were our primary concern.
I suspect it of being habit, which is what a lot of medical activity is
based
on, and much administrative activity. It is worth looking at habits to
see
if there is evidence of benefit from them.
So, _is_ the performance of the servers improved following a reboot?
Is the performance better on day 1 than on day 2?
If there is no detectable degradation in performance (which is what is
offered
as the problem avoided, rather than a sudden unpredictable catastrophic
stoppage with a blue screen) in day 2, might a cautious exension to day
3 4 5
6 7 and so on offer some information?
If the performance dips on day 7, then it would be nice to see why...
and
after such a long time of running, perhaps the fragmentation would be
measurable and the ownership of the segments could be determined.
--
Dr Adrian Midgley GP Exeter www.defoam.net
Open Source is a necessary but not of itself sufficient condition.
|