Charting server load

Thu Feb 1 19:00:00 UTC 2007

On 2/1/07, Ian Petersen <ispeters-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org> wrote:
>
> > That's quite helpful, thank you.  All of the machines that I run have
> > loads under 1, and so I have little experience with this type of thing.
>
> I believe, and someone can correct me if I'm wrong, that "maximum
> load" is the same as the number of CPUs in the system.  In other
> words, if you have a uniprocessor, and the load is 1, then the CPU is
> always busy, but you haven't overloaded the system.  In a
> dual-processor setup, you can run with a load of 2 and it's the same
> as a uniprocessor running at 1--both processors are maxed, but the
> system's not overloaded.

Ian's quite right .. depending on the version of top that's available, you
may get a lins starting with 'CPU states' for each processor, or you may
just get a single line starting with 'Cpu(s):'

A single processor machine might look like this:

 13:52:55  up 2 days,  7:28,  7 users,  load average: 0.07, 0.06, 0.09
89 processes: 88 sleeping, 1 running, 0 zombie, 0 stopped
CPU states:   2.8% user   0.4% system   0.0% nice   0.0% iowait  96.7% idle
Mem:   513772k av,  504120k used,    9652k free,       0k shrd,   82244k
buff
                    389064k actv,       0k in_d,   10468k in_c
Swap: 1044184k av,   13608k used, 1030576k free                  274352k
cached

  PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME CPU COMMAND
15726 alex      20   0  1144 1144   864 R     0.9  0.2   0:00   0 top
    1 root      15   0   480  452   424 S     0.0  0.0   0:04   0 init
    2 root      15   0     0    0     0 SW    0.0  0.0   0:00   0 keventd
    3 root      15   0     0    0     0 SW    0.0  0.0   0:01   0 kapmd
    4 root      34  19     0    0     0 SWN   0.0  0.0   0:00   0
ksoftirqd_CPU
    9 root      25   0     0    0     0 SW    0.0  0.0   0:00   0 bdflush
    5 root      15   0     0    0     0 SW    0.0  0.0   0:00   0 kswapd
    6 root      15   0     0    0     0 SW    0.0  0.0   0:00   0 kscand/DMA
...

and a multi-processor machine might look like this:

top - 13:53:14 up 389 days,  8:03,  1 user,  load average: 0.03, 0.01, 0.00
Tasks: 106 total,   1 running, 105 sleeping,   0 stopped,   0 zombie
Cpu(s):  0.0% us,  0.0% sy,  0.0% ni, 99.9% id,  0.0% wa,  0.0% hi,  0.0% si
Mem:   2074712k total,  1084732k used,   989980k free,   166684k buffers
Swap:  4194296k total,        0k used,  4194296k free,   359816k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+
COMMAND
12858 openvpn   15   0  7464 4536 1144 S  2.0  0.2  11:03.81openvpn
    1 root      16   0  1744  568  492 S  0.0  0.0   0:16.26init
    2 root      RT   0     0    0    0 S  0.0  0.0   0:15.90migration/0
    3 root      34  19     0    0    0 S  0.0  0.0   0:00.00ksoftirqd/0
    4 root      RT   0     0    0    0 S  0.0  0.0   0:09.78migration/1
    5 root      34  19     0    0    0 S  0.0  0.0   0:00.00ksoftirqd/1
    6 root      RT   0     0    0    0 S  0.0  0.0   0:07.55migration/2
    7 root      39  19     0    0    0 S  0.0  0.0   0:00.00ksoftirqd/2
    8 root      RT   0     0    0    0 S  0.0  0.0   0:16.08migration/3
    9 root      39  19     0    0    0 S  0.0  0.0   0:00.00ksoftirqd/3
   10 root      10  -5     0    0    0 S  0.0  0.0   0:14.95events/0
   11 root      10  -5     0    0    0 S  0.0  0.0   0:02.35events/1
   12 root      10  -5     0    0    0 S  0.0  0.0   0:01.45events/2
   13 root      10  -5     0    0    0 S  0.0  0.0   0:00.44events/3
   14 root      10  -5     0    0    0 S  0.0  0.0   0:00.01 khelper
...

Note that for the multi-processor machine there are multiple event queues.

So, assuming my understanding of the load number is correct, your web
> servers would need to be 40-processor machines to be able to run at a
> load of 40 without being overloaded.  Or, turning that on its head,
> your hosting company has apparently oversold the server by a factor of
> 10 to 40, depending on how many cores are in the machine, which is
> unreasonable in my eyes, unless its costing you pennies.

Right -- it may be that it's a dual or quad machine that's just keeping
busy, but my guess is that it's a single processor machine that the web
provider is making plenty of money from. Try the command

  $ cat /proc/cpuinfo

and see what it says -- it should list the available processors on the
system. The first system shown above has a single 1GHz processor with 256K
cache; the second one has four 2.4GHz processors, each with 512K cache.
Bogomips numbers are 2005 and 4799, respectively.

-- 
Alex Beamish
Toronto, Ontario
aka talexb
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gtalug.org/pipermail/legacy/attachments/20070201/332421bc/attachment.html>