next »

[nutballs]

alrighty...

I officially HATE linux's system usage reporting.

WTF is the important number for how much juice your server is pushing?
if you look in TOP, it reports the 1,5,15 minute load averages.
I also reports the %CPU.

I run at times (when i do major updating), a load of 4, but %CPU of only like 10-20%

I know that load is an indicator, and you shouldnt go over 1xprocessors.
But seriously, all my apps are millisecond responsive, even during the spike to 4.

can someone actually explain this damn thing in english, so it can be used to determine how much more headroom I have left, and can pump the box harder.... sheesh.

[vsloathe]

It's not really something you can just break down into English and then tack hard and fast rules to, at least not in my experience.

If you're not having performance problems, record the numbers and those are your baseline. If you start to have performance problems, look at the numbers and if they're way above your baseline, figure out what's causing them to be.

There's probably an easier way to do it but that's what I've always done. Then again I might be atypical among sysadmins for my general ataraxia towards performance and extreme pedanticism with regard to security.

[nutballs]

yea. thats what I feared.

I understand that the load average is actually based on processes, and god knows what else.
So you could have a huge load, like 40
but a cpu % of 5%

which if you ask me is retarded.
But hey, I guess I am spoiled by the magic metric you get in windows, which might be bullshit, but always seems to be spot on.
100% usage = nothing works
95% usage = stuff is slow.