You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mina.apache.org by Frédéric Brégier <fr...@free.fr> on 2007/01/18 20:50:00 UTC
Anyone get's experience with IBM JVM ?
Hi folks,
Does anyone get an experience with the JVM from IBM ?
I start to develop a quite complex project 6 months ago
using my laptop under Windows and a Sun JDK (1.5 32 bits).
This last two months I got no break inside my code,
even in heavy load. So I decides it was time to go to the next step.
I begin last week to make some benchmark
on a IBM PSeries (Power 5 cpu on AIX 5.3)
using IBM JDK (1.5 too but 64bits). I have no chance to get
a Sun JVM for AIX since it seems Sun is not releasing this kind of
JVM for AIX. And there is no IBM JVM for Windows
(in fact not really, but you need to have a IBM PC, which I don't have).
And I found some strange behaviour. Also, right now, I can't say
if this is my app or anything else (I use also Oracle 10g as part of
the component).
Some of the strange behaviour I found was for instance that I get
some null pointer exception on closing mina sessions, which I never
get before.
Again, I have not completely analyzed the problem so my point is :
Does someone have made some real work using the IBM JVM ?
I suspect different behaviour of JVMs...
Frederic
Re: Anyone get's experience with IBM JVM ?
Posted by Greg Duffy <gd...@gmail.com>.
While performance testing our app, I found that JRockit 1.5 took a while to
'warm up' (JIT optimization, I guess), while Sun 1.5 was faster than JRockit
only before the warm-up. However, Sun 1.6 seems to kick them both in the
pants while still lacking the long-ish warm-up time. I've never tried IBM's.
Anyway ... just another anecdote. :)
GD
On 1/18/07, Emmanuel Lecharny <el...@gmail.com> wrote:
>
> I have tested Apache DS with Sun JDK, IBM J9 and JRockit, either 1.4.x and
> 5.0 java VM.
>
> It was on intel processors, and Linux (FC5 an Ubuntu). So far, no problem.
> AIX is another beast...
>
> FYI, performance I got with IBM JVM was much better than with Sun JDK
> (around 30% better), and jrockit was even better (40%), but this is not
> all
> about MINA, even if MINA is the network layer of ADS.
>
> On 1/18/07, Frédéric Brégier <fr...@free.fr> wrote:
> >
> > Hi folks,
> >
> > Does anyone get an experience with the JVM from IBM ?
> >
> > I start to develop a quite complex project 6 months ago
> > using my laptop under Windows and a Sun JDK (1.5 32 bits).
> > This last two months I got no break inside my code,
> > even in heavy load. So I decides it was time to go to the next step.
> >
> > I begin last week to make some benchmark
> > on a IBM PSeries (Power 5 cpu on AIX 5.3)
> > using IBM JDK (1.5 too but 64bits). I have no chance to get
> > a Sun JVM for AIX since it seems Sun is not releasing this kind of
> > JVM for AIX. And there is no IBM JVM for Windows
> > (in fact not really, but you need to have a IBM PC, which I don't have).
> >
> > And I found some strange behaviour. Also, right now, I can't say
> > if this is my app or anything else (I use also Oracle 10g as part of
> > the component).
> > Some of the strange behaviour I found was for instance that I get
> > some null pointer exception on closing mina sessions, which I never
> > get before.
> >
> > Again, I have not completely analyzed the problem so my point is :
> >
> > Does someone have made some real work using the IBM JVM ?
> >
> > I suspect different behaviour of JVMs...
> >
> > Frederic
> >
>
>
>
> --
> Cordialement,
> Emmanuel Lécharny
> www.iktek.com
>
>
Re: Anyone get's experience with IBM JVM ?
Posted by Emmanuel Lecharny <el...@gmail.com>.
I have tested Apache DS with Sun JDK, IBM J9 and JRockit, either 1.4.x and
5.0 java VM.
It was on intel processors, and Linux (FC5 an Ubuntu). So far, no problem.
AIX is another beast...
FYI, performance I got with IBM JVM was much better than with Sun JDK
(around 30% better), and jrockit was even better (40%), but this is not all
about MINA, even if MINA is the network layer of ADS.
On 1/18/07, Frédéric Brégier <fr...@free.fr> wrote:
>
> Hi folks,
>
> Does anyone get an experience with the JVM from IBM ?
>
> I start to develop a quite complex project 6 months ago
> using my laptop under Windows and a Sun JDK (1.5 32 bits).
> This last two months I got no break inside my code,
> even in heavy load. So I decides it was time to go to the next step.
>
> I begin last week to make some benchmark
> on a IBM PSeries (Power 5 cpu on AIX 5.3)
> using IBM JDK (1.5 too but 64bits). I have no chance to get
> a Sun JVM for AIX since it seems Sun is not releasing this kind of
> JVM for AIX. And there is no IBM JVM for Windows
> (in fact not really, but you need to have a IBM PC, which I don't have).
>
> And I found some strange behaviour. Also, right now, I can't say
> if this is my app or anything else (I use also Oracle 10g as part of
> the component).
> Some of the strange behaviour I found was for instance that I get
> some null pointer exception on closing mina sessions, which I never
> get before.
>
> Again, I have not completely analyzed the problem so my point is :
>
> Does someone have made some real work using the IBM JVM ?
>
> I suspect different behaviour of JVMs...
>
> Frederic
>
--
Cordialement,
Emmanuel Lécharny
www.iktek.com
Re: Anyone get's experience with IBM JVM ?
Posted by Julien Vermillard <jv...@archean.fr>.
Le jeudi 18 janvier 2007 à 21:29 +0100, Frédéric Brégier a écrit :
> Thanks to point me the single / several CPU reasons.
> That could be the reason. So I need to re-check
> my code.
>
> Frederic
I tested java IBM 1.4 and 1.5 too, worked like a charm with MINA (i386
and AMD64 versions). Didn't tested CPU perfs, but the memory usage
wasn't much better than Sun JVM, so I choosed to keep Sun for now.
SMP systems are pretty good for showing you codec bugs, specialy if you
use a global one and not a codec per session, you should definitively
take a look at that.
Julien
PS : C'est sympa de voir que MINA commence à toucher Toulouse :)
Re: Anyone get's experience with IBM JVM ?
Posted by Frédéric Brégier <fr...@free.fr>.
Thanks to point me the single / several CPU reasons.
That could be the reason. So I need to re-check
my code.
Frederic
----- Original Message -----
> Does anyone get an experience with the JVM from IBM ?
Only a little. We tried it on linux (opteron). The IBM 1.5 exhibited
performance characteristics similar to the Sun JVM, and was quite a
bit slower than JRockit.
It did work though!
> And I found some strange behaviour. Also, right now, I can't say
> if this is my app or anything else (I use also Oracle 10g as part of
> the component).
> Some of the strange behaviour I found was for instance that I get
> some null pointer exception on closing mina sessions, which I never
> get before.
I'd be willing to bet that your PC is a single CPU and your AIX box
has several CPUs? It is probably a threading bug in your code. We
found that we could test our app on a single CPU box but only when we
moved it onto a real SMP box did all the threading bugs manifest
themselves...
We also found that the in-VM transport in MINA was great for
pinpointing subtle threading bugs.
RG
Re: Anyone get's experience with IBM JVM ?
Posted by Robert Greig <ro...@gmail.com>.
On 18/01/07, Frédéric Brégier <fr...@free.fr> wrote:
> Does anyone get an experience with the JVM from IBM ?
Only a little. We tried it on linux (opteron). The IBM 1.5 exhibited
performance characteristics similar to the Sun JVM, and was quite a
bit slower than JRockit.
It did work though!
> And I found some strange behaviour. Also, right now, I can't say
> if this is my app or anything else (I use also Oracle 10g as part of
> the component).
> Some of the strange behaviour I found was for instance that I get
> some null pointer exception on closing mina sessions, which I never
> get before.
I'd be willing to bet that your PC is a single CPU and your AIX box
has several CPUs? It is probably a threading bug in your code. We
found that we could test our app on a single CPU box but only when we
moved it onto a real SMP box did all the threading bugs manifest
themselves...
We also found that the in-VM transport in MINA was great for
pinpointing subtle threading bugs.
RG