News Group Posts

OSR5 double panic smp hyperthreading

p4

From: Bela Lubkin <belal@sco.com>
Subject: Re: Server crashes - need help! :(
Date: Thu, 2 Jan 2003 08:31:27 GMT
References: <3E0F2797.1040102@dniq-online.com>
<8cnCo4nVhFB@zocki.toppoint.de>
<3E0FDD8E.9090603@dniq-online.com>
<8crYG2XVhFB@zocki.toppoint.de>
<3E13BE3D.80006@dniq-online.com> 

Farlander wrote:

>    Load from phone calls hitting the server. The system takes up to 90% 
> of CPU. Here is the snapshot of 'sar' output on the server that has just 
> a few active sessions on it:
> 
> 22:03:48    %usr    %sys    %wio   %idle (-u)
> 22:03:49       8      67       2      24
> 22:03:50      16      67       0      17
> 22:03:51       3      42       0      55
> 22:03:52       8      62       3      27
> 
>    And here is the snapshot from another server - dual CPU - with twice 
> more active sessions:
> 
> 22:04:51    %usr    %sys    %wio   %idle (-u)
> 22:04:52       1      19       1      79
> 22:04:53       7      26       1      66
> 22:04:54       3      16       3      78
> 22:04:55       4      19       1      76
> 
>    There is some difference, no? :)

Please be more specific in these statements.  I think you mean "here is
a single P4 with a few sessions" and "here is a dual PIII with twice as
many sessions".  Put those words right inline so we know what you're
talking about all the way through.  Add in CPU clock speeds.  Slowest
P4's are 1.4GHz, fastest PIII's are close to that, this could be
perfectly reasonable (or close to it).  Sure, the _actual_ speeds are
probably farther apart, but if you don't spell it out we don't know.

OpenServer 5.0.6 prints a statement of clock speed at boot time.  It
also prints CPU info.  Like this:

  # egrep 'cpu|clock' /dev/string/cfg 
  %cpu      -               -   -  unit=1 family=6 type=Pentium III
  %cpuid    -               -   -  unit=1 vend=GenuineIntel tfms=0:6:8:10(2)
  %clock    -               -   -  type=TSC/999519129Hz
  %cpu      -             255   -  unit=2 family=6 type=Pentium III
  %cpuid    -               -   -  unit=2 vend=GenuineIntel tfms=0:6:8:6(2)

(from a dual Pentium III 1GHz machine).  Notice that this machine is
running different stepping CPUs (068A and 0686) -- generally not a good
idea, but seems to be working on this machine.

Show us your output from `egrep 'cpu|clock' /dev/string/cfg`, clearly
tagging each with your performance observations.  (It would probably
also be good to start a separate thread on this topic, something like
"P4 slower than expected".)

Also, run this ad hoc performance test on each:

  time awk 'BEGIN{j=10*1000*1000;for(i=0;i<j;i++)}'

I get about 3.25 seconds on a PIII-1GHz.  15.2 seconds on a Pentium Pro
200MHz.  Ignore the "real" and "sys" time, just look at "user".

Run it several times, while the system is under load and while it isn't.
Performance while not under load should be pretty constant (+/- a few
percent).  Under heavy load it might get 10-25% worse due to cache
sharing.

If performance varies by more than 25% from run to run, something is
wrong with the machine's memory subsystem.  Some areas of memory are not
cached, for instance.

>    That is possible. I was thinking about trying to do what was said 
> here before by Bela Lubkin - install SMP support and license on P4 in 
> order to activate its hyperthreading (if it supports one, which I, 
> honestly, don't know how to find out) and see if it's going to change 
> anything :)

Wow, that's an improbable option...  I have not seen a case where
Hyperthreading helps OSR5 performance (keeping in mind that so far, OSR5
has no clue about HT, does not know how to schedule processes to CPUs
properly to keep each hardware CPU busy).  If Rainer's idea is right, it
might actually help in this case.  One "CPU" could block in a device
driver while the other continued to run applications or other driver
code.  Interesting.

> > BTW(nothing to do with the panic):
> > Did you check the CPU heat?
> > If your boards generates such high loads and the cooling
> > is not good, the P4 will throttle itself...
> 
>    Would that cause output of HW command to change anyhow?

No, but it would definitely affect the outcome of the little `awk`
performance test.

>Bela<
 

Comments?

Jump to Comments

Many of the products and books I review are things I purchased for my own use. Some were given to me specifically for the purpose of reviewing them. I resell or can earn commissions from the sale of some of these items. Links within these pages may be affiliate links that pay me for referring you to them. That's mostly insignificant amounts of money; whenever it is not I have made my relationship plain. I also may own stock in companies mentioned here. If you have any question, please do feel free to contact me.

Specific links that take you to pages that allow you to purchase the item I reviewed are very likely to pay me a commission. Many of the books I review were given to me by the publishers specifically for the purpose of writing a review. These gifts and referral fees do not affect my opinions; I often give bad reviews anyway.

We use Google third-party advertising companies to serve ads when you visit our website. These companies may use information (not including your name, address, email address, or telephone number) about your visits to this and other websites in order to provide advertisements about goods and services of interest to you. If you would like more information about this practice and to know your choices about not having this information used by these companies, click here.