Date: Thu, 8 Jul 1999 18:52:09 -0700 (PDT) From: Doug <Doug@gorean.org> To: =?X-UNKNOWN?Q?Alejandro_Ram=EDrez?= <ales@megared.net.mx> Cc: freebsd-questions@FreeBSD.ORG Subject: Re: Uptime basics!!! Message-ID: <Pine.BSF.4.05.9907081840500.1046-100000@dt054n86.san.rr.com> In-Reply-To: <023b01bec99f$6da27920$fba3f9cf@megared.net.mx>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 8 Jul 1999, Alejandro Ram=EDrez wrote:
> Hi,
>=20
> I like to see the CPU usage at a given time, and when i write uptime
> command it gives me diferent values for 1,5 and 15 minutes average, but m=
y
> simple question is wich its the highest value that a CPU can have, becaus=
e I
> have notice that in sometimes it gives me 0.50, and in another times it
> givesme 2.35, and also I have seen numbers like 5.48 and I dont know how =
to
> interpret that in a % way.
=09The "highest" value is dependent on mostly hardware constraints,
and a few software things. I have some solaris servers that run with a
load average of 75. :) However, this makes more sense once you understand
what the numbers mean. There is no "percentage" involved, the number is
the number of processes in the queue waiting for a slice of CPU time. If
you have a load average of less than 1 it generally means that every
process is getting as much cpu as it needs as soon as it needs it. If your
load average is more than one, it means that there is at least one process
that is always waiting for cpu time. Whether this is acceptable to you or
not is purely a subjective measurement. If your system is "fast enough"
for you, then your load average is fine. If your system is sluggish and
you notice that your load average is 2 or 5 (or 75) then you have to take
steps to correct it.=20
=09Depending on what you are measuring the percentage of idle cpu
displayed in 'top' output is usually more significant, however this
depends on the process. A heavily loaded IRC server for example actually
becomes *more* efficient as it approaches certain peaks on various
resource utilization curves. For instance, I had a server that would
regularly run 5,000 clients or more. The first 2,000 loaded the cpu from
about 0% to about 80%. The next 2,000 went from 80% to 99% or more, but
once it hit its stride I could fit another 1,500 clients on without
breaking a sweat. Unfortunately our ircd code sucked, so we peaked out
around there. However some efnet servers running an almost identical OS
configuration and better ircd code eventually were able to break my
record, then smashed it to bits with about 8,000 simultaneous connections.=
=20
=09Hmmm.. I had a point here somewhere... oh yes. My point is that
depending on what you are measuring there probably isn't going to be a
nice, tidy number to represent it. Welcome to unix system administration.
Doug
--=20
On account of being a democracy and run by the people, we are the only
nation in the world that has to keep a government four years, no matter
what it does.
-- Will Rogers
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.05.9907081840500.1046-100000>