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

lars lars at gmx.at
Thu Feb 16 13:52:56 PST 2006

Eric Schultz wrote:
> lars wrote:
>> A long uptime means that the machine hasn't been rebooted for a long
>> time. If that time's longer than the time to the last patch that
>> required a kernel recompilation and a reboot, it means the server is not
>> patched. Where's the point in advertising an unpatched machine?
> Good afternoon...
> Perhaps it means the OS doesn't need to be patched that frequently
Possibly. But patch frequency means what exactly?

> or has a patch mechanism that avoids reboots?  
Exactly, that doesn't exist (yet).

Although there was something in a Usenix proceeding or somewhere else,
about "micro-reboots" where, to use FreeBSD wordage,
Base and Ports' programs where so modularised to allow this.
Thus making only, say, a driver or some kernel component reboot,
but the majority of the system stays up.
Of course a reboot of the NIC's driver kills that component's "uptime".

> That's certainly worth advertising (if only were true).
Actually it (this website) means advertising an unpatched machine 
running unpatched services not available to the outside.

> The top machine has been running for almost 6 years on FreeBSD 3.3 which 
> means the admin probably believes that "if it ain't broke, don't fix 
> it."
Which is not necessarily the best strategy.
But may be right in this case.

> I would also want to advertise the longevity of an OS.
You mean the ability of that OS to run so long without requiring a reboot?
I'm not sure that's that relevant nowadays.
How many OS aren't capable of staying up long?
Even Windows doesn't need too much Viagra to keep it up.

> (You might not like that last one if you're a hardware vendor :)
> Also, a lot of work-arounds for security patches amount to "lock the 
> front door."  
What do you mean by that?

> So perhaps some systems don't need to be patched because 
> they're administered so as not to require immediate patching/upgrading.
If your machine only runs an NFS daemon and is behind a firewall,
ok, you don't need to patch it asap when an NFS SA and patch is issued, 
if all clients connecting to the machine are benign.

I could also run a machine in some private net protected by firewalls
and whatnot running only this uptime program.
Unless I lose power or some hardware failure occurs I'll have a long
uptime. A bit useless though.

I think that 'uptime' and this website fail to define precisely enough
what the point of the exercise is to be able to make useful conclusions
about something about some OS.

What exactly do you want to measure to make what decision?

	Do you want to find out how much [%] your OS is available
	whithout load just patching it with the latest SA recommended 			patches?

	Do you want to find out how much [%] your OS is available
	[can serve 100 FTP users simultaneously at wire speed
	with this NIC]
	just patching it with the latest SA recommended patches?

	Do you want see how long an unpatched OS version can keep it up
	without any patches or interaction whatsoever?


More information about the freebsd-questions mailing list