We recently purchased couple of T5200 from Sun. We have Java J2EE
Application server (WebLogic) running on them and the OS is Solaris
10. Each server has 4 instances of WebLogic running on them. Since
5200 has 64 threads (which I think is overkill), is it possible to
break down 64 threads such that the OS gets 16 and each of the
WebLogic instances gets 12 (so basically 16+12+12+12+12) ?

I have read about processor sets and containers but I am note very
clear on the implementation. Not sure what is the right way to
approach this problem. I just want input from others to see what
approach they have taken with this setup.

Thanks,
Kevin.

Kevin wrote:
> We recently purchased couple of T5200 from Sun. We have Java J2EE
> Application server (WebLogic) running on them and the OS is Solaris
> 10. Each server has 4 instances of WebLogic running on them. Since
> 5200 has 64 threads (which I think is overkill), is it possible to
> break down 64 threads such that the OS gets 16 and each of the
> WebLogic instances gets 12 (so basically 16+12+12+12+12) ?

Why you want to give 15 threads to the OS ? And why divide the whole 
number of threads into a subsets (12) ? Why don't you trust OS scheduler
and let it decide how to spread all working processes on all threads ?

If it is reasonable create 4 zones and put each WebLogic into it.
And then, if you have performance problems, try to optimize using
processor sets.
At the very begin just start with default.

przemol
--
http://przemol.blogspot.com/

In article <ae5e9bbf-faaa-4efa-bab4-db439c8cd9f5@y22g2000prd.googlegroups.com>,
	Kevin <kejoseph@hotmail.com> writes:
> We recently purchased couple of T5200 from Sun. We have Java J2EE
> Application server (WebLogic) running on them and the OS is Solaris
> 10. Each server has 4 instances of WebLogic running on them. Since
> 5200 has 64 threads (which I think is overkill), is it possible to
> break down 64 threads such that the OS gets 16 and each of the
> WebLogic instances gets 12 (so basically 16+12+12+12+12) ?
> 
> I have read about processor sets and containers but I am note very
> clear on the implementation. Not sure what is the right way to
> approach this problem. I just want input from others to see what
> approach they have taken with this setup.

Yes you can do this, and sometimes you will be required to do so
because an application doesn't play well with sharing the system,
or because different teams are funding the service and want some
guaranteed chunk of it reserved for themselves, or because you
will be loading more applications onto the system and you don't
want the initial ones to get a misleading impression of what level
of service they will see long-term, etc. You also need to decide
if you want to impose a max number of CPUs to be available to an
application (even if the system has some idle) or a min number
reserved which can be exceed this if the system is otherwise idle.

However, before going down this road, I suggest you just try running
with all the Solaris defaults. Otherwise you might simply be over-
complicating the system configuration for no observable benefit, or
possibly quite some detriment as your scheduling contraints may be
less optimal than Solaris does by default. Obviously, this depends
on the nature of your applications and what type of response they
are required to provide, and I am not personally familiar with
configuring WebLogic.

-- 
Andrew Gabriel
[email address is not usable -- followup in the newsgroup]

On Jun 28, 2:53=A0pm, Kevin <kejos...@hotmail.com> wrote:
> We recently purchased couple of T5200 from Sun. We have Java J2EE
> Application server (WebLogic) running on them and the OS is Solaris
> 10. Each server has 4 instances of WebLogic running on them. Since
> 5200 has 64 threads (which I think is overkill), is it possible to
> break down 64 threads such that the OS gets 16 and each of the
> WebLogic instances gets 12 (so basically 16+12+12+12+12) ?
>
> I have read about processor sets and containers but I am note very
> clear on the implementation. Not sure what is the right way to
> approach this problem. I just want input from others to see what
> approach they have taken with this setup.
>
> Thanks,
> Kevin.

Take a look here:  http://www.sun.com/software/solaris/howtoguides/containe=
rsLowRes.js

The answer really depends on your requirements, with the solution
often using a combination of features.  I prefer not to use processor
sets, because of the potential for wasted CPU cycles.  Using the fair
share scheduler (FSS), you can assign CPU shares to projects and
zones.  FSS will make all CPU cycles available to any process on the
system as long as there is no resource contention.  Once there IS
resource contention, FSS kicks in to keep processes in check according
to the assigned number of shares.

Kevin schrieb:
> We recently purchased couple of T5200 from Sun. We have Java J2EE
> Application server (WebLogic) running on them and the OS is Solaris
> 10. Each server has 4 instances of WebLogic running on them. Since
> 5200 has 64 threads (which I think is overkill), is it possible to
> break down 64 threads such that the OS gets 16 and each of the
> WebLogic instances gets 12 (so basically 16+12+12+12+12) ?
> 
> I have read about processor sets and containers but I am note very
> clear on the implementation. Not sure what is the right way to
> approach this problem. I just want input from others to see what
> approach they have taken with this setup.
> 
> Thanks,
> Kevin.
> 

with version of WLS you run? I heard in some older versions (6 or 7 I 
think), that it does not scale well in systems with more than 4 CPUs. 
This time the Sparc Processor with 2 threads came out and doubles the 
"CPUs" per board.

there are some migration documents for the niagara systems, if i 
remember correctly also for J2EE Appl. Servers.

W.

WebLogic 9.2 - I dont think that can be an issue here.