You are viewing a plain text version of this content. The canonical link for it is here.
Posted to j-users@xalan.apache.org by nareshk <na...@virtusa.com> on 2010/10/22 11:19:34 UTC

Xalan-J 2.7.1 is 3-4 times slower on unix

I am running SimpleTransform that is part of examples in a loop for 20 times
and measuring the performance of transform method. It is taking 3-4 times
more on Unix than in windows. I would assume that it will be much faster on
a server class machine, however the reverse is true. Any body has any ideas
why this is happening

JDK 1.5.0_25 on both solaris server and windows machines.
-- 
View this message in context: http://old.nabble.com/Xalan-J-2.7.1-is-3-4-times-slower-on-unix-tp30026977p30026977.html
Sent from the Xalan - J - Users mailing list archive at Nabble.com.


Re: Xalan-J 2.7.1 is 3-4 times slower on unix

Posted by Claus Kick <cl...@googlemail.com>.
Hello Naresh,

the machine does not have 64 CPUs. It has up to two CPUs with 8 cores
each with 8 threads each.
Judging from your machine, you have one CPU with 8 cores with 8
threads (or two CPUs with 4 cores with 8 threads each).

What is more, raw CPU speed ranges from 1.2 Ghz to 1.6 Ghz, something
each common machine beats hands down in raw speed.

Have a look here:

https://globalsp.ts.fujitsu.com/dmsp/docs/ds-se-t5240-en.pdf

The virtualisation does not help with speed, as it eats a little bit
of performance for each action which takes place.

So, this does not have anything to do with xalan or even Solaris. It
is merely a hardware thing.

It is hard to get a T Series machine to a grinding halt where the
system does not react any longer due to the CPU architecture.
But the tradeoff is that the machine is not very fast for each
individual action.

2010/10/27 Naresh Kumar Kolli <Na...@virtusa.com>:
> We are running on a soalris virtualized machine. It is Sunw-T5240 series with 64 CPUs.
> I am not 100% sure how much is allocated to current slice.
>
> I would think this should have a better through put. Can you explain if possible.
>
> Thanks & Regards.
> Naresh Kolli
>
>
> -----Original Message-----
> From: Claus Kick [mailto:claus.kick@googlemail.com]
> Sent: Tuesday, October 26, 2010 8:00 PM
> To: Naresh Kumar Kolli
> Subject: Re: Xalan-J 2.7.1 is 3-4 times slower on unix
>
> Ok, what sun, sorry oracle hardware does it run on?
> It if is one of the T-series, it is understandable, as per core
> throughput is not as high as for a high end single-core or less-core
> machine.
>
> 2010/10/25 Naresh Kumar Kolli <Na...@virtusa.com>:
>> I am not measuring any startup time at all. I am only measuring the transformation time which is 3-4 times slower on solaris.
>>
>> It has the same behavior if I use -server or -client option on solaris. I am also using xalan-2.7.1 and xerces2.9.0.
>>
>> Thanks & Regards.
>> Naresh
>>
>> -----Original Message-----
>> From: Claus Kick [mailto:claus.kick@googlemail.com]
>> Sent: Saturday, October 23, 2010 8:28 PM
>> To: Naresh Kumar Kolli
>> Subject: Re: Xalan-J 2.7.1 is 3-4 times slower on unix
>>
>> 2010/10/22 nareshk <na...@virtusa.com>
>>>
>>> I am running SimpleTransform that is part of examples in a loop for 20 times
>>> and measuring the performance of transform method. It is taking 3-4 times
>>> more on Unix than in windows. I would assume that it will be much faster on
>>> a server class machine, however the reverse is true. Any body has any ideas
>>> why this is happening
>>>
>>> JDK 1.5.0_25 on both solaris server and windows machines.
>>>
>>
>> How is the JVM configured on the solaris machine? There are a few
>> options which definitely influence jvm startup time.
>>
>> Virtusa was recently ranked and featured in 2010 Global Services 100, IAOP's 2010 Global Outsourcing 100 sub-list, 2009 Deloitte Technology Fast 500 and 2009 Dataquest-IDC Best Employers Survey among others.
>>
>> --------------------------------------------------------------------------------------------
>>
>> This message, including any attachments, contains confidential information
>> intended for a specific individual and purpose, and is intended for the addressee only. Any unauthorized disclosure, use, dissemination, copying, or distribution of
>> this message or any of its attachments or the information contained in this e-mail, or the taking of any action based on it, is strictly prohibited. If you are not the intended recipient, please notify the sender immediately by return e-mail and delete this message.
>>
>> --------------------------------------------------------------------------------------------
>>
>
> Virtusa was recently ranked and featured in 2010 Global Services 100, IAOP's 2010 Global Outsourcing 100 sub-list, 2009 Deloitte Technology Fast 500 and 2009 Dataquest-IDC Best Employers Survey among others.
>
> --------------------------------------------------------------------------------------------
>
> This message, including any attachments, contains confidential information
> intended for a specific individual and purpose, and is intended for the addressee only. Any unauthorized disclosure, use, dissemination, copying, or distribution of
> this message or any of its attachments or the information contained in this e-mail, or the taking of any action based on it, is strictly prohibited. If you are not the intended recipient, please notify the sender immediately by return e-mail and delete this message.
>
> --------------------------------------------------------------------------------------------
>