Hercules on 8 CPUs
kvshetye
Mar 28, 2007
Hi all,
Today I will be installing hercules on a 8 way(processor) linux
box. Dont know what MIPS I should get on that. But will update you
all as soon as I finish that installation and configuration.
CPU details are: (from /proc/cpuinfo)
Processor : 0
.....
processor : 7
vendor_id : GenuineIntel
cpu family : 15
model : 2
model name : Intel(R) XEON(TM) MP CPU 1.90GHz
stepping : 2
cpu MHz : 2121.068
cache size : 1024 KB
fdiv_bug : no
hlt_bug : no
f00f_bug : no
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 2
wp : yes
flags : fpu tsc msr pae mce cx8 apic mtrr mca cmov pat
pse36 clflush d
ts acpi mmx fxsr sse sse2 ss ht tm up
bogomips : 4755.03
I guess it is a 4 way system with all DUO/Hperthreading processors.
OS Details: (uname -a)
Linux xxx.xx.xxx.com 2.6.18-8.el5xen #1 SMP Fri Jan 26 14:42:21 EST
2007 i686 i6
86 i386 GNU/Linux
Regards,
Kailas
5:48 am
Re: Hercules on 8 CPUs
shupe7
Apr 7, 2007
I'll be curious to hear how you make out with that. Some time ago, I
installed Herc on a two-way P4 Xeon 2.8GHz box. I was surprised to
find that it did not run much better (or indicate many more MIPS)
than running on an AMD Athlon XP 2400+ uniprocessor. So I just kept
Herc on the AMD machine and used the Intel for a way-overpriced
workstation. It was kind of a bummer, though, because I'd bought the
dual processor machine specifically for Herc and blew a bunch of $$$
on it. Oh well, live & learn...
Good luck with yours.
-Shupe7
--- In hercules-390@yahoogroups.com, "kvshetye" <kvshetye@...> wrote:
>
> Hi all,
>
> Today I will be installing hercules on a 8 way(processor) linux
> box. Dont know what MIPS I should get on that. But will update
you
> all as soon as I finish that installation and configuration.
>
> CPU details are: (from /proc/cpuinfo)
>
> Processor : 0
> .....
> processor : 7
> vendor_id : GenuineIntel
> cpu family : 15
> model : 2
> model name : Intel(R) XEON(TM) MP CPU 1.90GHz
> stepping : 2
> cpu MHz : 2121.068
> cache size : 1024 KB
> fdiv_bug : no
> hlt_bug : no
> f00f_bug : no
> coma_bug : no
> fpu : yes
> fpu_exception : yes
> cpuid level : 2
> wp : yes
> flags : fpu tsc msr pae mce cx8 apic mtrr mca cmov pat
> pse36 clflush d
> ts acpi mmx fxsr sse sse2 ss ht tm up
> bogomips : 4755.03
>
> I guess it is a 4 way system with all DUO/Hperthreading processors.
>
> OS Details: (uname -a)
>
> Linux xxx.xx.xxx.com 2.6.18-8.el5xen #1 SMP Fri Jan 26 14:42:21 EST
> 2007 i686 i6
> 86 i386 GNU/Linux
>
> Regards,
>
> Kailas
>
8:15 pm
Re: Hercules on 8 CPUs
David Wade
Apr 7, 2007
I am surprised that you donā€™t much difference. I guess it depends on
workload..
Dave Wade
Illegitimi Non Carborundum
-----Original Message-----
From: hercules-390@yahoogroups.com [mailto:hercules-390@yahoogroups.com] On
Behalf Of shupe7
Sent: 07 April 2007 21:17
To: hercules-390@yahoogroups.com
Subject: [hercules-390] Re: Hercules on 8 CPUs
I'll be curious to hear how you make out with that. Some time ago, I
installed Herc on a two-way P4 Xeon 2.8GHz box. I was surprised to
find that it did not run much better (or indicate many more MIPS)
than running on an AMD Athlon XP 2400+ uniprocessor. So I just kept
Herc on the AMD machine and used the Intel for a way-overpriced
workstation. It was kind of a bummer, though, because I'd bought the
dual processor machine specifically for Herc and blew a bunch of $$$
on it. Oh well, live & learn...
Good luck with yours.
-Shupe7
--- In hercules-390@ <mailto:hercules-390%40yahoogroups.com> yahoogroups.com,
"kvshetye" <kvshetye@...> wrote:
>
> Hi all,
>
> Today I will be installing hercules on a 8 way(processor) linux
> box. Dont know what MIPS I should get on that. But will update
you
> all as soon as I finish that installation and configuration.
>
> CPU details are: (from /proc/cpuinfo)
>
> Processor : 0
> .....
> processor : 7
> vendor_id : GenuineIntel
> cpu family : 15
> model : 2
> model name : Intel(R) XEON(TM) MP CPU 1.90GHz
> stepping : 2
> cpu MHz : 2121.068
> cache size : 1024 KB
> fdiv_bug : no
> hlt_bug : no
> f00f_bug : no
> coma_bug : no
> fpu : yes
> fpu_exception : yes
> cpuid level : 2
> wp : yes
> flags : fpu tsc msr pae mce cx8 apic mtrr mca cmov pat
> pse36 clflush d
> ts acpi mmx fxsr sse sse2 ss ht tm up
> bogomips : 4755.03
>
> I guess it is a 4 way system with all DUO/Hperthreading processors.
>
> OS Details: (uname -a)
>
> Linux xxx.xx.xxx.com 2.6.18-8.el5xen #1 SMP Fri Jan 26 14:42:21 EST
> 2007 i686 i6
> 86 i386 GNU/Linux
>
> Regards,
>
> Kailas
>
[Non-text portions of this message have been removed]
9:40 pm
Re: Hercules on 8 CPUs
Enrico Sorichetti
Apr 8, 2007
--- In hercules-390@yahoogroups.com, "David Wade" <g4ugm@...> wrote:
>
> I am surprised that you donā€™t much difference. I guess it depends on
workload..
>
>
>
> Dave Wade
> Illegitimi Non Carborundum
>
I wouold be surprised if You had :-)
We should make clear what we/You are trying to measure,
MIPS ( meaningless indicator of processor speed )
IMHO the number of processors available is ininfluent for a MIPS oriented
benchmark
there would be no I/O involved so the single processor speed would be the only
affecting
factor
Thruput benchmark ...
here the multitasking approach of hercules might help;
( different threads for I/O processing )
but.... since we will not be able to emulate a full fledged CPU environment
( let's say 5000 cics/ims terminals, 200 transactions per second, 30 cics
regions, 10 db2
subsystems, 20 active batch initiators... add anything I overlooked )
then I tend to believe that anything more then a dual environment would be kind
of
overshooting.
Just an opinion ....
regards
enrico sorichetti
7:09 am
Re: Hercules on 8 CPUs
kvshetye
Apr 9, 2007
Hi,
You are correct. Not getting the expected increase in performance.
I was expecting IPL time to be reduced. But it did not happen. I
get more speed with my T60 (Intel Duo Core).
Planning to create a test screnario for this test speed which will
have something like you mentioned. Will update you what happens
when more than 1000 CICS terminals connect and use psuedo-
conversational programs on it. Not planning to use more than two
cics regions.
I tested time of IPL procedure on various machines and found that
the best performance we get with Duo processors with NUMCPU=2. May
be hercules/zos is spending a lot of time in synchronizing the CPUs
than doing the actual job in case of more than two CPUs.
thanks and regards,
Kailas Shetye
--- In hercules-390@yahoogroups.com, "Enrico Sorichetti"
<e.sorichetti@...> wrote:
>
> --- In hercules-390@yahoogroups.com, "David Wade" <g4ugm@> wrote:
> >
> > I am surprised that you donā€™t much difference. I guess it
depends on workload..
> >
> >
> >
> > Dave Wade
> > Illegitimi Non Carborundum
> >
>
> I wouold be surprised if You had :-)
>
> We should make clear what we/You are trying to measure,
>
> MIPS ( meaningless indicator of processor speed )
> IMHO the number of processors available is ininfluent for a MIPS
oriented benchmark
> there would be no I/O involved so the single processor speed would
be the only affecting
> factor
>
>
> Thruput benchmark ...
> here the multitasking approach of hercules might help;
> ( different threads for I/O processing )
> but.... since we will not be able to emulate a full fledged CPU
environment
> ( let's say 5000 cics/ims terminals, 200 transactions per
second, 30 cics regions, 10 db2
> subsystems, 20 active batch initiators... add anything I
overlooked )
>
> then I tend to believe that anything more then a dual environment
would be kind of
> overshooting.
>
> Just an opinion ....
>
> regards
>
> enrico sorichetti
>
5:44 am
Re: Hercules on 8 CPUs
Jay Maynard
Apr 9, 2007
On Mon, Apr 09, 2007 at 05:44:46AM -0000, kvshetye wrote:
> You are correct. Not getting the expected increase in performance.
> I was expecting IPL time to be reduced. But it did not happen. I
> get more speed with my T60 (Intel Duo Core).
On a system with a lot of processors such as the one you're describing, your
performance is going to depend on how well the guest OS can exploit multiple
CPUs. I don't know how well the common OSes do that, especially during IPL.
Hercules will run multiple CPUs at the expected multiplier of a single CPU's
speed, less whatever overhead is mandated by the architecture. The best case
is nothing requiring multiple CPU synchronization, of course, and that shows
an almost linear improvement. My development box got upgraded not too long
ago to a 4-core (dual Opteron 275) configuration, and it turns 355 MIPS in a
tight loop on four emulated processors. That's not a very useful number for
the real world, of course, but it sounds nice. :-) Real-world numbers are
closer to 60 MIPS. I haven't timed such things as IPLing OSes; I should
probably do that next time I'm home.
--
Jay Maynard, K5ZC http://www.conmicro.cx
http://jmaynard.livejournal.com http://www.tronguy.net
http://www.hercules-390.org (Yes, that's me!)
Buy Hercules stuff at http://www.cafepress.com/hercules-390
8:10 am
Copyright 2007