PDA

View Full Version : Provider-1 and multi-processor machine



cciesec2006
2008-02-06, 21:43
I have Provider-1 NG AI R55 w/ HFA_20 Manager + Container running on
SPLAT on a Dell P4 1.3 GHz quad processors with 2GB RAM. I have a
single CMA running on this box. It seems like P-1 is not taking advantage
of multi-processors box. When I run "top" I see this:

8:33pm up 57 min, 1 user, load average: 1.00, 1.01, 0.77
56 processes: 54 sleeping, 2 running, 0 zombie, 0 stopped
CPU0 states: 0.1% user, 3.0% system, 0.0% nice, 95.0% idle
CPU1 states: 0.0% user, 0.0% system, 0.0% nice, 100.0% idle
CPU2 states: 0.0% user, 0.0% system, 0.0% nice, 100.0% idle
CPU3 states: 99.0% user, 0.1% system, 0.0% nice, 0.0% idle
Mem: 2059224K av, 2048540K used, 10684K free, 312K shrd, 41656K buff
Swap: 4192944K av, 0K used, 4192944K free 1838276K cached

As you can see CPU3 is running at 100% utilization while the other 3 CPUs
is at 100% idle.

Does it mean that Provider-1 NG with AI R55 can not take advantage
of multi-processors machine?

Am I wrong?

chillyjim
2008-02-07, 11:51
That's pretty much the case. It is highly dependent on the OS scheduler to assign processes to the CPUs. Seeing as most of the management features are user-space programs, I would have expected to see somewhat better process distribution though.

mamakos
2008-02-12, 18:17
On Solaris, it does take advantage of every processors...

chillyjim
2008-02-12, 18:27
Solaris has a better scheduler. That's one of the reasons a lot of P1 customers are still using Sun/Solaris for management.

cciesec2006
2008-02-13, 00:22
Solaris has a better scheduler. That's one of the reasons a lot of P1 customers are still using Sun/Solaris for management.

I have to disagree on this. One of the MANY reasons people are moving away
from running P-1 on Solaris and moving toward P-1 on SPLAT is because
Checkpoint TAC support for P-1 on Sun Solaris has been nothing but
an embarrassment. Checkpoint TAC will find anything possible to blame
the issue on Solaris OS. I've lived through this nightmare for almost five
years when we first rolled out P-1 NG Feature Pack 3 (multiple
managers and multiple containers). I remembered that I had an issue
about MDS in "collision" with our P-1 infrastructure. I opened a TAC
case with Checkpoint and that TAC case lasted almost 14 months
without any resolution. Checkpoint blamed the problem that we run
Veritas backup on the Solaris Manager(s) box. I eventually closed the
case and the issue remained. Fortunately, I've moved on to another
job. I have no doubt that those MDSes will crash and burn hard.

When you run P-1 on SPLAT and you have issues, open a ticket with
CP and they are responsible for it. They can't finger-pointing anywhere
else. Oh wait, I guess they can still blame the hardware as the root cause.

my 2c.