From: alexd on
[third time lucky. news.aioe.org has accepted the previous 2 attempts to
post this, but they don't appear]

Load average out of peak time is about 1. 'vmstat 30' shows about 30%
user, 5% sys, 60% idle and 5% wait.

It does seem to be a bit of a beast, but the experience for the end user
is determined by the Javascript performance of their browser or other
MUA if they're not using the web interface.

Of course there's nothing stopping you sticking it in a VM and playing
with it for yourself. It's assembled mostly from open source components,
so there's plenty of scope for tinkering. If it's in a VM you could
presumably screw it right down so the VM only gets a small portion of
host CPU.

Off on a tangent, what you want is Martin Gregorie's frugal email DB,
with Roundcube as a front end and some sort of admin tool for managing
email addresses and accounts.

--
<http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm(a)ale.cx)
15:48:41 up 51 days, 19:09, 3 users, load average: 0.03, 0.15, 0.28
It is better to have been wasted and then sober
than to never have been wasted at all
From: Theo Markettos on
alexd <troffasky(a)hotmail.com> wrote:
> Load average out of peak time is about 1. 'vmstat 30' shows about 30%
> user, 5% sys, 60% idle and 5% wait.

Hmmm... so load average 1 but 60% idle?

> Of course there's nothing stopping you sticking it in a VM and playing
> with it for yourself. It's assembled mostly from open source components,
> so there's plenty of scope for tinkering. If it's in a VM you could
> presumably screw it right down so the VM only gets a small portion of
> host CPU.
>
> Off on a tangent, what you want is Martin Gregorie's frugal email DB,
> with Roundcube as a front end and some sort of admin tool for managing
> email addresses and accounts.

Thanks. I'll have a look.

Theo
From: Gordon Henderson on
In article <o+f*0yu7s(a)news.chiark.greenend.org.uk>,
Theo Markettos <theom+news(a)chiark.greenend.org.uk> wrote:
>alexd <troffasky(a)hotmail.com> wrote:
>> Load average out of peak time is about 1. 'vmstat 30' shows about 30%
>> user, 5% sys, 60% idle and 5% wait.
>
>Hmmm... so load average 1 but 60% idle?

Load Average is not always a good indication of what a server is
doing. It's a rather crude metric (IMO) - handy at times though. Basically
it indicates the number of processes that are currently running - but,
a running process could be waiting on hardware, thus not using any CPU
cycles! I've seen badly behaved systems with load averages over 100,
yet still be fast and responsive to stuff cached in RAM.

The sys and wait numbers would interest me here - it seems to be spending
time waiting on (disk?) IO and in the kernel - either task switching or
dealing with IO requests.

If you run ps or top, look for processes in a D state - that's stopped,
but waiting on IO which will contribute to "load".

Gordon
From: alexd on
On 02/04/10 08:32, Gordon Henderson wrote:
> In article<o+f*0yu7s(a)news.chiark.greenend.org.uk>,
> Theo Markettos<theom+news(a)chiark.greenend.org.uk> wrote:
>> alexd<troffasky(a)hotmail.com> wrote:
>>> Load average out of peak time is about 1. 'vmstat 30' shows about 30%
>>> user, 5% sys, 60% idle and 5% wait.
>>
>> Hmmm... so load average 1 but 60% idle?

If you under-resource a VM [for example], load average will shoot up
whilst it appears to be doing nothing.

> The sys and wait numbers would interest me here - it seems to be spending
> time waiting on (disk?) IO and in the kernel - either task switching or
> dealing with IO requests.

Yes, I'm pretty sure it's IO.

I'll post a few lines of vmstat output at some point, to satisfy the
pent up curiosity of all.

--
<http://ale.cx/> (AIM:troffasky) (UnSoEsNpEaTm(a)ale.cx)
08:43:23 up 57 days, 13:22, 1 user, load average: 0.03, 0.11, 0.09
It is better to have been wasted and then sober
than to never have been wasted at all