XANSYS Forum Index
    Register    

FAQ    Search    Memberlist    Usergroups    SmartFeedSmartFeed    Profile    Log in
[Xansys] Model solution time
 
This forum is locked: you cannot post, reply to, or edit topics.   This topic is locked: you cannot edit posts or make replies.    XANSYS Forum Index -> XANSYS
Author Message
mohammad.gharaibeh
User


Joined: 01 Nov 2012
Posts: 132
Location: Zarqa, Jordan

PostPosted: Tue May 26, 2015 5:55 pm  Reply with quote

Dear XANSYS Members,

I have a general question regarding the solution time of a nonlinear
simulation.

I have a relatively large model (800K elements) and I'm running a nonlinear
static analysis. The nonlinearity comes from the material I'm using,
hyper-elastic material model. I used to run this model successfully with
approximate solution time of 6 hours. Now, the same model takes about 13
hours to solve and I'm not sure why. Keeping in mind that I use the same
HPC settings.

Is there any reason of this? One important thing I should mention is that I
run my models over the server provided by my school and not on a standalone
computer.

Any explanations would be greatly appreciated.

Thanks,
Mohammad


Mohammad A Gharaibeh
Binghamton University


--
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
christopher.wright
User


Joined: 17 Jun 2009
Posts: 927

PostPosted: Tue May 26, 2015 6:44 pm  Reply with quote

On May 26, 2015, at 7:55 PM, Mohammad A Gharaibeh wrote:

Quote:
One important thing I should mention is that I
run my models over the server provided by my school and not on a standalone
computer.

Check the wall clock (elapsed) time against the actual solution time. If the solution time hasn't changed all that much, while the wall clock time has changed, you're not being allowed access to processing cycles for some reason. Could be someone just has a higher priority than you do or that your memory allotment is low adn your job is 'thrashing.'

Christopher Wright P.E. |"They couldn't hit an elephant at
chrisw@skypoint.com | this distance" (last words of Gen.
.......................................| John Sedgwick, Spotsylvania 1864)
http://www.skypoint.com/members/chrisw/



+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
saad.bin.aziz
User


Joined: 28 Aug 2014
Posts: 19

PostPosted: Tue May 26, 2015 6:59 pm  Reply with quote

Dear Mr.Gharaibeh,

What is the main difference between your two simulation? can you pls
clarify.Was your first run on a PC and the second run on HPC?
I faced the same situation in HPC last year. Later on I found that in HPC i
need to assign more node to make it quicker.

Thanks
Saad Aziz
PhD Student
Louisiana State University


On Tue, May 26, 2015 at 7:55 PM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Dear XANSYS Members,

I have a general question regarding the solution time of a nonlinear
simulation.

I have a relatively large model (800K elements) and I'm running a nonlinear
static analysis. The nonlinearity comes from the material I'm using,
hyper-elastic material model. I used to run this model successfully with
approximate solution time of 6 hours. Now, the same model takes about 13
hours to solve and I'm not sure why. Keeping in mind that I use the same
HPC settings.

Is there any reason of this? One important thing I should mention is that I
run my models over the server provided by my school and not on a standalone
computer.

Any explanations would be greatly appreciated.

Thanks,
Mohammad


Mohammad A Gharaibeh
Binghamton University


--
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
kenneth.gomez
Site Admin


Joined: 14 May 2012
Posts: 16
Location: Perth, WA, Australia

PostPosted: Wed May 27, 2015 2:21 am  Reply with quote

Mohammad,

If you can show us the time summary output from both runs, that would help.

Running on a server, in addition to what Christopher has said, also depends on the system resources. You could have been lucky and had your first job run fully in RAM but in the second instance, the RAM was fully utilized and your job was swapping onto the hard disk, which reaaaaaaly slows down the job.

How many other students also use the server? I presume the server has a batching software, like PBS or similar?

Kenneth Gomez
Senior Engineering Specialist - Subsea, INTECSEA
Tel+61 (0)8 9289 5858


-----Original Message-----
From: Xansys [mailto:xansys-bounces@xansys.org] On Behalf Of Christopher Wright
Sent: Wednesday, 27 May 2015 9:44 AM
To: ANSYS User Discussion List
Subject: Re: [Xansys] Model solution time


On May 26, 2015, at 7:55 PM, Mohammad A Gharaibeh wrote:

Quote:
One important thing I should mention is that I run my models over the
server provided by my school and not on a standalone computer.

Check the wall clock (elapsed) time against the actual solution time. If the solution time hasn't changed all that much, while the wall clock time has changed, you're not being allowed access to processing cycles for some reason. Could be someone just has a higher priority than you do or that your memory allotment is low adn your job is 'thrashing.'

Christopher Wright P.E. |"They couldn't hit an elephant at
chrisw@skypoint.com | this distance" (last words of Gen.
.......................................| John Sedgwick, Spotsylvania 1864) http://www.skypoint.com/members/chrisw/



+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+
*** WORLEYPARSONS GROUP NOTICE *** "This email is confidential. If you are not the intended recipient, you must not disclose or use the information contained in it. If you have received this email in error, please notify us immediately by return email and delete the email and any attachments. Any personal views or opinions expressed by the writer may not necessarily reflect the views or opinions of any company in the WorleyParsons Group of Companies."
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
_________________
Best regards,
Kenneth Gomez
Senior Engineering Specialist (Subsea Pipelines)
INTECSEA (Perth)
Back to top
View user's profile Send private message Visit poster's website
mohammad.gharaibeh
User


Joined: 01 Nov 2012
Posts: 132
Location: Zarqa, Jordan

PostPosted: Wed May 27, 2015 7:09 am  Reply with quote

Dear Kenneth, Saad and Chris,

Thank you all for your replies. I highly appreciate that.

Kenneth,

Quote:
Quote:
How many other students also use the server? I presume the server has a
batching software, like PBS or similar?

As I know, there are many students who works on ANSYS, about 12 students.
But, I quite don't understand what you mean by batching software, PBS? Can
you explain further, please?


I'm attaching the time summary from both runs, the quick and slow.


*QUICK>>*
+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/12/2015 Time: 20:34
Windows Process ID: 1356

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build: 40219)
Intel(R) Math Kernel Library Version 11.0.5 Product Build 20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\LargeDie\300_330

Total CPU time for main thread : 14464.3 seconds
Total CPU time summed for all threads : 143189.4 seconds

Elapsed time spent pre-processing model (/PREP7) : 89.6 seconds
Elapsed time spent solution - preprocessing : 15.0 seconds
Elapsed time spent computing solution : 20171.9 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 4612.9 seconds

Equation solver computational rate : 31709.5 Mflops
Equation solver effective I/O rate : 7038.1 MB/sec

Maximum total memory used : 22655.0 MB
Maximum total memory allocated : 23808.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+



!
***********************************************************************************************!
!
***********************************************************************************************!

*SLOW>>*

+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/27/2015 Time: 09:46
Windows Process ID: 1476

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build: 40219)
Intel(R) Math Kernel Library Version 11.0.5 Product Build 20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\SmallDie\125_155

Total CPU time for main thread : 84237.7 seconds
Total CPU time summed for all threads : 948167.7 seconds

Elapsed time spent pre-processing model (/PREP7) : 214.5 seconds
Elapsed time spent solution - preprocessing : 41.5 seconds
Elapsed time spent computing solution : 124945.4 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 35800.0 seconds

Equation solver computational rate : 37654.2 Mflops
Equation solver effective I/O rate : 6614.6 MB/sec

Maximum total memory used : 33127.0 MB
Maximum total memory allocated : 36706.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+


I can see that the times in the slow model are higher than those of the
quick one. Also, the memory usage in the slow is larger than that of slow
model. My question is, what controls the memory usage? Can I so something
for it?

Your suggestions are highly appreciated.

Thanks,
Mohammad
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
saad.bin.aziz
User


Joined: 28 Aug 2014
Posts: 19

PostPosted: Wed May 27, 2015 7:27 am  Reply with quote

Dear Mr. Gharaibeh,

I had a quick look at your two file comparison, i think there is a
significant difference in */POST1, quick model (Elapsed time spent
post-processing model (/POST1) : 4612.9 seconds) and slow model(Elapsed
time spent post-processing model (/POST1) : 35800.0 seconds).*

Pls check your slow model, my guess, there are lot of things you assigned
in /POST1 section.

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:09 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Dear Kenneth, Saad and Chris,

Thank you all for your replies. I highly appreciate that.

Kenneth,

Quote:
Quote:
How many other students also use the server? I presume the server has a
batching software, like PBS or similar?

As I know, there are many students who works on ANSYS, about 12 students.
But, I quite don't understand what you mean by batching software, PBS? Can
you explain further, please?


I'm attaching the time summary from both runs, the quick and slow.


*QUICK>>*
+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/12/2015 Time: 20:34
Windows Process ID: 1356

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build: 40219)
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\LargeDie\300_330

Total CPU time for main thread : 14464.3 seconds
Total CPU time summed for all threads : 143189.4 seconds

Elapsed time spent pre-processing model (/PREP7) : 89.6 seconds
Elapsed time spent solution - preprocessing : 15.0 seconds
Elapsed time spent computing solution : 20171.9 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 4612.9 seconds

Equation solver computational rate : 31709.5 Mflops
Equation solver effective I/O rate : 7038.1 MB/sec

Maximum total memory used : 22655.0 MB
Maximum total memory allocated : 23808.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+



!

***********************************************************************************************!
!

***********************************************************************************************!

*SLOW>>*

+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/27/2015 Time: 09:46
Windows Process ID: 1476

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build: 40219)
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\SmallDie\125_155

Total CPU time for main thread : 84237.7 seconds
Total CPU time summed for all threads : 948167.7 seconds

Elapsed time spent pre-processing model (/PREP7) : 214.5 seconds
Elapsed time spent solution - preprocessing : 41.5 seconds
Elapsed time spent computing solution : 124945.4 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 35800.0 seconds

Equation solver computational rate : 37654.2 Mflops
Equation solver effective I/O rate : 6614.6 MB/sec

Maximum total memory used : 33127.0 MB
Maximum total memory allocated : 36706.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+


I can see that the times in the slow model are higher than those of the
quick one. Also, the memory usage in the slow is larger than that of slow
model. My question is, what controls the memory usage? Can I so something
for it?

Your suggestions are highly appreciated.

Thanks,
Mohammad
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
mohammad.gharaibeh
User


Joined: 01 Nov 2012
Posts: 132
Location: Zarqa, Jordan

PostPosted: Wed May 27, 2015 7:37 am  Reply with quote

Hi Saad,

Thanks for looking into my output files.

Yes, you are right. There is a significant difference between quick and
slow model in POST1. That was my question, why slow model is too slow?

In /POST1, I compute volume averaged stresses in a specific set of elements
then I write them into an ASCII file.

Also, how about solution time? Quick model: 20171.9 seconds. The Slow
model: 124945.4 seconds. Isn't that very significant?

Thank you for your time and suggestions.

Mohammad

--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================


On Wed, May 27, 2015 at 10:27 AM, Saad Aziz <saziz1@tigers.lsu.edu> wrote:

Quote:
Dear Mr. Gharaibeh,

I had a quick look at your two file comparison, i think there is a
significant difference in */POST1, quick model (Elapsed time spent
post-processing model (/POST1) : 4612.9 seconds) and slow model(Elapsed
time spent post-processing model (/POST1) : 35800.0 seconds).*

Pls check your slow model, my guess, there are lot of things you assigned
in /POST1 section.

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:09 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Dear Kenneth, Saad and Chris,

Thank you all for your replies. I highly appreciate that.

Kenneth,

Quote:
Quote:
How many other students also use the server? I presume the server has a
batching software, like PBS or similar?

As I know, there are many students who works on ANSYS, about 12 students.
But, I quite don't understand what you mean by batching software, PBS?
Can
Quote:
you explain further, please?


I'm attaching the time summary from both runs, the quick and slow.


*QUICK>>*
+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/12/2015 Time: 20:34
Windows Process ID: 1356

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build: 40219)
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\LargeDie\300_330

Total CPU time for main thread : 14464.3 seconds
Total CPU time summed for all threads : 143189.4 seconds

Elapsed time spent pre-processing model (/PREP7) : 89.6 seconds
Elapsed time spent solution - preprocessing : 15.0 seconds
Elapsed time spent computing solution : 20171.9 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 4612.9 seconds

Equation solver computational rate : 31709.5 Mflops
Equation solver effective I/O rate : 7038.1 MB/sec

Maximum total memory used : 22655.0 MB
Maximum total memory allocated : 23808.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+



!


***********************************************************************************************!
Quote:
!


***********************************************************************************************!
Quote:

*SLOW>>*

+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/27/2015 Time: 09:46
Windows Process ID: 1476

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build: 40219)
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\SmallDie\125_155

Total CPU time for main thread : 84237.7 seconds
Total CPU time summed for all threads : 948167.7 seconds

Elapsed time spent pre-processing model (/PREP7) : 214.5 seconds
Elapsed time spent solution - preprocessing : 41.5 seconds
Elapsed time spent computing solution : 124945.4 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 35800.0 seconds

Equation solver computational rate : 37654.2 Mflops
Equation solver effective I/O rate : 6614.6 MB/sec

Maximum total memory used : 33127.0 MB
Maximum total memory allocated : 36706.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+


I can see that the times in the slow model are higher than those of the
quick one. Also, the memory usage in the slow is larger than that of slow
model. My question is, what controls the memory usage? Can I so something
for it?

Your suggestions are highly appreciated.

Thanks,
Mohammad
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
edward.carman
User


Joined: 21 Oct 2008
Posts: 59
Location: Johannesburg, South Africa

PostPosted: Wed May 27, 2015 7:42 am  Reply with quote

Are your models actually the same? There are significant time increases at
all stages from pre-processing to post-processing, while the solver
computation rate and I/O rate are roughly the same for each run. This
indicates that the solving machine itself is probably not the problem.

Looking at working directory names, it also looks like you may be running
different model configurations. How many equilibrium iterations did each
model take to solve and how many nodes did each model have? What solver
were you using in each case?

Finally, I see that you are running the Shared Memory Parallel (SMP) solver
on 12 cores. Generally, around 4 cores and higher you should consider using
the distributed solver (DMP) and you will often see performance reduction
when running SMP on more than 8 cores. This is covered in the Performance
Guide in the MAPDL manual (help/ans_per/per_book.html)

--
Edward Carman
Technical Services
ANSYS UK Ltd
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
saad.bin.aziz
User


Joined: 28 Aug 2014
Posts: 19

PostPosted: Wed May 27, 2015 7:50 am  Reply with quote

Dear Mr. Gharaibeh,

yes there was also time gap in solution time also.

1. does your quick model has more element than slow model?

2. Also, I agree with Mr. Edward's comment-
"Finally, I see that you are running the Shared Memory Parallel (SMP) solver
on 12 cores. Generally, around 4 cores and higher you should consider using
the distributed solver (DMP) and you will often see performance reduction
when running SMP on more than 8 cores. This is covered in the Performance
Guide in the MAPDL manual (help/ans_per/per_book.html)"

My suggestion, use less core. In my case I had quicker solution time by
using one core consist of 20 processor rather than 4 core consist of 20
processor each(4*20=80 processor).

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:36 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Hi Saad,

Thanks for looking into my output files.

Yes, you are right. There is a significant difference between quick and
slow model in POST1. That was my question, why slow model is too slow?

In /POST1, I compute volume averaged stresses in a specific set of elements
then I write them into an ASCII file.

Also, how about solution time? Quick model: 20171.9 seconds. The Slow
model: 124945.4 seconds. Isn't that very significant?

Thank you for your time and suggestions.

Mohammad

--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================


On Wed, May 27, 2015 at 10:27 AM, Saad Aziz <saziz1@tigers.lsu.edu> wrote:

Quote:
Dear Mr. Gharaibeh,

I had a quick look at your two file comparison, i think there is a
significant difference in */POST1, quick model (Elapsed time spent
post-processing model (/POST1) : 4612.9 seconds) and slow
model(Elapsed
Quote:
time spent post-processing model (/POST1) : 35800.0 seconds).*

Pls check your slow model, my guess, there are lot of things you assigned
in /POST1 section.

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:09 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Dear Kenneth, Saad and Chris,

Thank you all for your replies. I highly appreciate that.

Kenneth,

Quote:
Quote:
How many other students also use the server? I presume the server
has a
Quote:
Quote:
batching software, like PBS or similar?

As I know, there are many students who works on ANSYS, about 12
students.
Quote:
Quote:
But, I quite don't understand what you mean by batching software, PBS?
Can
Quote:
you explain further, please?


I'm attaching the time summary from both runs, the quick and slow.


*QUICK>>*
+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/12/2015 Time: 20:34
Windows Process ID: 1356

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build:
40219)
Quote:
Quote:
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\LargeDie\300_330

Total CPU time for main thread : 14464.3 seconds
Total CPU time summed for all threads : 143189.4 seconds

Elapsed time spent pre-processing model (/PREP7) : 89.6 seconds
Elapsed time spent solution - preprocessing : 15.0 seconds
Elapsed time spent computing solution : 20171.9 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 4612.9 seconds

Equation solver computational rate : 31709.5 Mflops
Equation solver effective I/O rate : 7038.1 MB/sec

Maximum total memory used : 22655.0 MB
Maximum total memory allocated : 23808.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+



!



***********************************************************************************************!
Quote:
Quote:
!



***********************************************************************************************!
Quote:
Quote:

*SLOW>>*

+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/27/2015 Time: 09:46
Windows Process ID: 1476

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build:
40219)
Quote:
Quote:
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\SmallDie\125_155

Total CPU time for main thread : 84237.7 seconds
Total CPU time summed for all threads : 948167.7 seconds

Elapsed time spent pre-processing model (/PREP7) : 214.5 seconds
Elapsed time spent solution - preprocessing : 41.5 seconds
Elapsed time spent computing solution : 124945.4 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 35800.0 seconds

Equation solver computational rate : 37654.2 Mflops
Equation solver effective I/O rate : 6614.6 MB/sec

Maximum total memory used : 33127.0 MB
Maximum total memory allocated : 36706.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+


I can see that the times in the slow model are higher than those of the
quick one. Also, the memory usage in the slow is larger than that of
slow
Quote:
Quote:
model. My question is, what controls the memory usage? Can I so
something
Quote:
Quote:
for it?

Your suggestions are highly appreciated.

Thanks,
Mohammad
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
peter.attar
User


Joined: 21 Oct 2008
Posts: 67

PostPosted: Wed May 27, 2015 8:03 am  Reply with quote

If your model is exactly the same, as you seem to have stated in your first post, than Chris Wright's answer was the correct one and you need to do a better job of understanding your school's HPC structure..that is not a job for the list.

I will say quickly that when you run on a cluster with a batch queuing system, you have the ability to ask for your job to be exclusive on any compute nodes which you are assigned to. In other words your job is the only one which being run on a set of processors. This may nor may not be applicable to you depending on your HPC environment.


Peter Attar
Associate Professor
The University of Oklahoma
School of Aerospace and Mechanical Engineering
865 Asp Ave.
Felgar Hall Room 212
Norman, OK 73019-1052
phone:405-325-1749
fax:405-325-1088

________________________________________
From: Xansys [xansys-bounces@xansys.org] on behalf of Saad Aziz [saziz1@tigers.lsu.edu]
Sent: Wednesday, May 27, 2015 9:50 AM
To: ANSYS User Discussion List
Subject: Re: [Xansys] Model solution time

Dear Mr. Gharaibeh,

yes there was also time gap in solution time also.

1. does your quick model has more element than slow model?

2. Also, I agree with Mr. Edward's comment-
"Finally, I see that you are running the Shared Memory Parallel (SMP) solver
on 12 cores. Generally, around 4 cores and higher you should consider using
the distributed solver (DMP) and you will often see performance reduction
when running SMP on more than 8 cores. This is covered in the Performance
Guide in the MAPDL manual (help/ans_per/per_book.html)"

My suggestion, use less core. In my case I had quicker solution time by
using one core consist of 20 processor rather than 4 core consist of 20
processor each(4*20=80 processor).

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:36 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Hi Saad,

Thanks for looking into my output files.

Yes, you are right. There is a significant difference between quick and
slow model in POST1. That was my question, why slow model is too slow?

In /POST1, I compute volume averaged stresses in a specific set of elements
then I write them into an ASCII file.

Also, how about solution time? Quick model: 20171.9 seconds. The Slow
model: 124945.4 seconds. Isn't that very significant?

Thank you for your time and suggestions.

Mohammad

--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================


On Wed, May 27, 2015 at 10:27 AM, Saad Aziz <saziz1@tigers.lsu.edu> wrote:

Quote:
Dear Mr. Gharaibeh,

I had a quick look at your two file comparison, i think there is a
significant difference in */POST1, quick model (Elapsed time spent
post-processing model (/POST1) : 4612.9 seconds) and slow
model(Elapsed
Quote:
time spent post-processing model (/POST1) : 35800.0 seconds).*

Pls check your slow model, my guess, there are lot of things you assigned
in /POST1 section.

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:09 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Dear Kenneth, Saad and Chris,

Thank you all for your replies. I highly appreciate that.

Kenneth,

Quote:
Quote:
How many other students also use the server? I presume the server
has a
Quote:
Quote:
batching software, like PBS or similar?

As I know, there are many students who works on ANSYS, about 12
students.
Quote:
Quote:
But, I quite don't understand what you mean by batching software, PBS?
Can
Quote:
you explain further, please?


I'm attaching the time summary from both runs, the quick and slow.


*QUICK>>*
+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/12/2015 Time: 20:34
Windows Process ID: 1356

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build:
40219)
Quote:
Quote:
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\LargeDie\300_330

Total CPU time for main thread : 14464.3 seconds
Total CPU time summed for all threads : 143189.4 seconds

Elapsed time spent pre-processing model (/PREP7) : 89.6 seconds
Elapsed time spent solution - preprocessing : 15.0 seconds
Elapsed time spent computing solution : 20171.9 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 4612.9 seconds

Equation solver computational rate : 31709.5 Mflops
Equation solver effective I/O rate : 7038.1 MB/sec

Maximum total memory used : 22655.0 MB
Maximum total memory allocated : 23808.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+



!



***********************************************************************************************!
Quote:
Quote:
!



***********************************************************************************************!
Quote:
Quote:

*SLOW>>*

+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/27/2015 Time: 09:46
Windows Process ID: 1476

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build:
40219)
Quote:
Quote:
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\SmallDie\125_155

Total CPU time for main thread : 84237.7 seconds
Total CPU time summed for all threads : 948167.7 seconds

Elapsed time spent pre-processing model (/PREP7) : 214.5 seconds
Elapsed time spent solution - preprocessing : 41.5 seconds
Elapsed time spent computing solution : 124945.4 seconds
Elapsed time spent solution - postprocessing : 0.0 seconds
Elapsed time spent post-processing model (/POST1) : 35800.0 seconds

Equation solver computational rate : 37654.2 Mflops
Equation solver effective I/O rate : 6614.6 MB/sec

Maximum total memory used : 33127.0 MB
Maximum total memory allocated : 36706.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+


I can see that the times in the slow model are higher than those of the
quick one. Also, the memory usage in the slow is larger than that of
slow
Quote:
Quote:
model. My question is, what controls the memory usage? Can I so
something
Quote:
Quote:
for it?

Your suggestions are highly appreciated.

Thanks,
Mohammad
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
mohammad.gharaibeh
User


Joined: 01 Nov 2012
Posts: 132
Location: Zarqa, Jordan

PostPosted: Wed May 27, 2015 11:22 am  Reply with quote

Hi Edward,

Thanks for your comments.

Quote:
Quote:
Are your models actually the same? There are significant time increases at
all stages from pre-processing to post-processing, while the solver
computation rate and I/O rate are roughly the same for each run. This
indicates that the solving machine itself is probably not the problem.

Both models have the same number of elements and nodes. The only difference
that a geometric parameter of a specific part is different in both models.
I didn't expect this change will have big effect on solution time.


Quote:
Quote:
Looking at working directory names, it also looks like you may be running
different model configurations. How many equilibrium iterations did each
model take to solve and how many nodes did each model have? What solver
were you using in each case?

Yes, different configurations as I mentioned above. In both models, I have
2 loadsteps with 8 substeps each. I use sparse solver.


Quote:
Quote:
Finally, I see that you are running the Shared Memory Parallel (SMP)
solver
Quote:
Quote:
on 12 cores. Generally, around 4 cores and higher you should consider
using
Quote:
Quote:
the distributed solver (DMP) and you will often see performance reduction
when running SMP on more than 8 cores. This is covered in the Performance
Guide in the MAPDL manual (help/ans_per/per_book.html)

Yes, I'm using SMP with 12 cores. I will look into DMP or (MPP as I checked
in the Launcher). Thanks for pointing out on this.


Thank you very much,
Mohammad
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================


On Wed, May 27, 2015 at 10:41 AM, Edward Carman <edward.carman@ansys.com>
wrote:

Quote:
Are your models actually the same? There are significant time increases at
all stages from pre-processing to post-processing, while the solver
computation rate and I/O rate are roughly the same for each run. This
indicates that the solving machine itself is probably not the problem.

Looking at working directory names, it also looks like you may be running
different model configurations. How many equilibrium iterations did each
model take to solve and how many nodes did each model have? What solver
were you using in each case?

Finally, I see that you are running the Shared Memory Parallel (SMP) solver
on 12 cores. Generally, around 4 cores and higher you should consider using
the distributed solver (DMP) and you will often see performance reduction
when running SMP on more than 8 cores. This is covered in the Performance
Guide in the MAPDL manual (help/ans_per/per_book.html)

--
Edward Carman
Technical Services
ANSYS UK Ltd
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
mohammad.gharaibeh
User


Joined: 01 Nov 2012
Posts: 132
Location: Zarqa, Jordan

PostPosted: Wed May 27, 2015 11:24 am  Reply with quote

Hi Saad,

Yes, both models have the same number of elements and nodes.

Quote:
Quote:
My suggestion, use less core. In my case I had quicker solution time by
using one core consist of 20 processor rather than 4 core consist of 20
processor each(4*20=80 processor).

I will look deeply into this and see if this improves things.

Thank you very much.

--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================


On Wed, May 27, 2015 at 10:50 AM, Saad Aziz <saziz1@tigers.lsu.edu> wrote:

Quote:
Dear Mr. Gharaibeh,

yes there was also time gap in solution time also.

1. does your quick model has more element than slow model?

2. Also, I agree with Mr. Edward's comment-
"Finally, I see that you are running the Shared Memory Parallel (SMP)
solver
on 12 cores. Generally, around 4 cores and higher you should consider using
the distributed solver (DMP) and you will often see performance reduction
when running SMP on more than 8 cores. This is covered in the Performance
Guide in the MAPDL manual (help/ans_per/per_book.html)"

My suggestion, use less core. In my case I had quicker solution time by
using one core consist of 20 processor rather than 4 core consist of 20
processor each(4*20=80 processor).

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:36 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Hi Saad,

Thanks for looking into my output files.

Yes, you are right. There is a significant difference between quick and
slow model in POST1. That was my question, why slow model is too slow?

In /POST1, I compute volume averaged stresses in a specific set of
elements
Quote:
then I write them into an ASCII file.

Also, how about solution time? Quick model: 20171.9 seconds. The Slow
model: 124945.4 seconds. Isn't that very significant?

Thank you for your time and suggestions.

Mohammad

--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================


On Wed, May 27, 2015 at 10:27 AM, Saad Aziz <saziz1@tigers.lsu.edu>
wrote:
Quote:

Quote:
Dear Mr. Gharaibeh,

I had a quick look at your two file comparison, i think there is a
significant difference in */POST1, quick model (Elapsed time spent
post-processing model (/POST1) : 4612.9 seconds) and slow
model(Elapsed
Quote:
time spent post-processing model (/POST1) : 35800.0 seconds).*

Pls check your slow model, my guess, there are lot of things you
assigned
Quote:
Quote:
in /POST1 section.

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:09 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Dear Kenneth, Saad and Chris,

Thank you all for your replies. I highly appreciate that.

Kenneth,

Quote:
Quote:
How many other students also use the server? I presume the server
has a
Quote:
Quote:
batching software, like PBS or similar?

As I know, there are many students who works on ANSYS, about 12
students.
Quote:
Quote:
But, I quite don't understand what you mean by batching software,
PBS?
Quote:
Quote:
Can
Quote:
you explain further, please?


I'm attaching the time summary from both runs, the quick and slow.


*QUICK>>*
+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/12/2015 Time: 20:34
Windows Process ID: 1356

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build:
40219)
Quote:
Quote:
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\LargeDie\300_330

Total CPU time for main thread : 14464.3
seconds
Quote:
Quote:
Quote:
Total CPU time summed for all threads : 143189.4
seconds
Quote:
Quote:
Quote:

Elapsed time spent pre-processing model (/PREP7) : 89.6
seconds
Quote:
Quote:
Quote:
Elapsed time spent solution - preprocessing : 15.0
seconds
Quote:
Quote:
Quote:
Elapsed time spent computing solution : 20171.9
seconds
Quote:
Quote:
Quote:
Elapsed time spent solution - postprocessing : 0.0
seconds
Quote:
Quote:
Quote:
Elapsed time spent post-processing model (/POST1) : 4612.9
seconds
Quote:
Quote:
Quote:

Equation solver computational rate : 31709.5 Mflops
Equation solver effective I/O rate : 7038.1 MB/sec

Maximum total memory used : 22655.0 MB
Maximum total memory allocated : 23808.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+



!




***********************************************************************************************!
Quote:
Quote:
Quote:
!




***********************************************************************************************!
Quote:
Quote:
Quote:

*SLOW>>*

+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/27/2015 Time: 09:46
Windows Process ID: 1476

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build:
40219)
Quote:
Quote:
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\SmallDie\125_155

Total CPU time for main thread : 84237.7
seconds
Quote:
Quote:
Quote:
Total CPU time summed for all threads : 948167.7
seconds
Quote:
Quote:
Quote:

Elapsed time spent pre-processing model (/PREP7) : 214.5
seconds
Quote:
Quote:
Quote:
Elapsed time spent solution - preprocessing : 41.5
seconds
Quote:
Quote:
Quote:
Elapsed time spent computing solution : 124945.4
seconds
Quote:
Quote:
Quote:
Elapsed time spent solution - postprocessing : 0.0
seconds
Quote:
Quote:
Quote:
Elapsed time spent post-processing model (/POST1) : 35800.0
seconds
Quote:
Quote:
Quote:

Equation solver computational rate : 37654.2 Mflops
Equation solver effective I/O rate : 6614.6 MB/sec

Maximum total memory used : 33127.0 MB
Maximum total memory allocated : 36706.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+


I can see that the times in the slow model are higher than those of
the
Quote:
Quote:
Quote:
quick one. Also, the memory usage in the slow is larger than that of
slow
Quote:
Quote:
model. My question is, what controls the memory usage? Can I so
something
Quote:
Quote:
for it?

Your suggestions are highly appreciated.

Thanks,
Mohammad
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
mohammad.gharaibeh
User


Joined: 01 Nov 2012
Posts: 132
Location: Zarqa, Jordan

PostPosted: Wed May 27, 2015 11:28 am  Reply with quote

Hi Peter,

On Wed, May 27, 2015 at 11:02 AM, Attar, Peter J. <peter.attar@ou.edu>
wrote:

Quote:
If your model is exactly the same, as you seem to have stated in your
first post, than Chris Wright's answer was the correct one and you need to
do a better job of understanding your school's HPC structure..that is not a
job for the list.


What I meant by the same is that both models have the same number of
elements and same number of loadsteps and substeps. Also, I didn't ask
anyone in the list to undestand the HPC structure of my school for me. Did
I? I will talk to the IT guys to understand things better.


Thanks for your suggestions.

Regards,
Mohammad

Note - Sorry for sending many emails. But I wanted to answer everyone in a
separate email. Thank you all for your patience.
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================


On Wed, May 27, 2015 at 11:02 AM, Attar, Peter J. <peter.attar@ou.edu>
wrote:

Quote:
If your model is exactly the same, as you seem to have stated in your
first post, than Chris Wright's answer was the correct one and you need to
do a better job of understanding your school's HPC structure..that is not a
job for the list.

I will say quickly that when you run on a cluster with a batch queuing
system, you have the ability to ask for your job to be exclusive on any
compute nodes which you are assigned to. In other words your job is the
only one which being run on a set of processors. This may nor may not be
applicable to you depending on your HPC environment.


Peter Attar
Associate Professor
The University of Oklahoma
School of Aerospace and Mechanical Engineering
865 Asp Ave.
Felgar Hall Room 212
Norman, OK 73019-1052
phone:405-325-1749
fax:405-325-1088

________________________________________
From: Xansys [xansys-bounces@xansys.org] on behalf of Saad Aziz [
saziz1@tigers.lsu.edu]
Sent: Wednesday, May 27, 2015 9:50 AM
To: ANSYS User Discussion List
Subject: Re: [Xansys] Model solution time

Dear Mr. Gharaibeh,

yes there was also time gap in solution time also.

1. does your quick model has more element than slow model?

2. Also, I agree with Mr. Edward's comment-
"Finally, I see that you are running the Shared Memory Parallel (SMP)
solver
on 12 cores. Generally, around 4 cores and higher you should consider using
the distributed solver (DMP) and you will often see performance reduction
when running SMP on more than 8 cores. This is covered in the Performance
Guide in the MAPDL manual (help/ans_per/per_book.html)"

My suggestion, use less core. In my case I had quicker solution time by
using one core consist of 20 processor rather than 4 core consist of 20
processor each(4*20=80 processor).

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:36 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Hi Saad,

Thanks for looking into my output files.

Yes, you are right. There is a significant difference between quick and
slow model in POST1. That was my question, why slow model is too slow?

In /POST1, I compute volume averaged stresses in a specific set of
elements
Quote:
then I write them into an ASCII file.

Also, how about solution time? Quick model: 20171.9 seconds. The Slow
model: 124945.4 seconds. Isn't that very significant?

Thank you for your time and suggestions.

Mohammad

--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================


On Wed, May 27, 2015 at 10:27 AM, Saad Aziz <saziz1@tigers.lsu.edu>
wrote:
Quote:

Quote:
Dear Mr. Gharaibeh,

I had a quick look at your two file comparison, i think there is a
significant difference in */POST1, quick model (Elapsed time spent
post-processing model (/POST1) : 4612.9 seconds) and slow
model(Elapsed
Quote:
time spent post-processing model (/POST1) : 35800.0 seconds).*

Pls check your slow model, my guess, there are lot of things you
assigned
Quote:
Quote:
in /POST1 section.

Regards,

Saad Aziz
PhD Student
Louisiana State University

On Wed, May 27, 2015 at 9:09 AM, Mohammad A Gharaibeh <
mgharai1@binghamton.edu> wrote:

Quote:
Dear Kenneth, Saad and Chris,

Thank you all for your replies. I highly appreciate that.

Kenneth,

Quote:
Quote:
How many other students also use the server? I presume the server
has a
Quote:
Quote:
batching software, like PBS or similar?

As I know, there are many students who works on ANSYS, about 12
students.
Quote:
Quote:
But, I quite don't understand what you mean by batching software,
PBS?
Quote:
Quote:
Can
Quote:
you explain further, please?


I'm attaching the time summary from both runs, the quick and slow.


*QUICK>>*
+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/12/2015 Time: 20:34
Windows Process ID: 1356

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build:
40219)
Quote:
Quote:
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\LargeDie\300_330

Total CPU time for main thread : 14464.3
seconds
Quote:
Quote:
Quote:
Total CPU time summed for all threads : 143189.4
seconds
Quote:
Quote:
Quote:

Elapsed time spent pre-processing model (/PREP7) : 89.6
seconds
Quote:
Quote:
Quote:
Elapsed time spent solution - preprocessing : 15.0
seconds
Quote:
Quote:
Quote:
Elapsed time spent computing solution : 20171.9
seconds
Quote:
Quote:
Quote:
Elapsed time spent solution - postprocessing : 0.0
seconds
Quote:
Quote:
Quote:
Elapsed time spent post-processing model (/POST1) : 4612.9
seconds
Quote:
Quote:
Quote:

Equation solver computational rate : 31709.5 Mflops
Equation solver effective I/O rate : 7038.1 MB/sec

Maximum total memory used : 22655.0 MB
Maximum total memory allocated : 23808.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+



!




***********************************************************************************************!
Quote:
Quote:
Quote:
!




***********************************************************************************************!
Quote:
Quote:
Quote:

*SLOW>>*

+--------------------- A N S Y S S T A T I S T I C S
------------------------+

Release: 15.0 Build: UP20131014 Platform: WINDOWS x64
Date Run: 05/27/2015 Time: 09:46
Windows Process ID: 1476

Processor Model: Intel(R) Xeon(R) CPU X5670 @ 2.93GHz

Compiler: Intel(R) FORTRAN Compiler Version 12.1.0 (Build: 20120928)
Microsoft(R) Visual C/C++ Compiler Version 16.0 (Build:
40219)
Quote:
Quote:
Intel(R) Math Kernel Library Version 11.0.5 Product Build
20130612

Total number of cores available : 24
Number of physical cores available : 12
Number of cores requested : 12 (Shared Memory Parallel)

GPU Acceleration: Not Requested

Job Name: Solar
Working Directory: Y:\Mohammad\ANSYS
Workspace\HeatSink\HyperElatic\SmallDie\125_155

Total CPU time for main thread : 84237.7
seconds
Quote:
Quote:
Quote:
Total CPU time summed for all threads : 948167.7
seconds
Quote:
Quote:
Quote:

Elapsed time spent pre-processing model (/PREP7) : 214.5
seconds
Quote:
Quote:
Quote:
Elapsed time spent solution - preprocessing : 41.5
seconds
Quote:
Quote:
Quote:
Elapsed time spent computing solution : 124945.4
seconds
Quote:
Quote:
Quote:
Elapsed time spent solution - postprocessing : 0.0
seconds
Quote:
Quote:
Quote:
Elapsed time spent post-processing model (/POST1) : 35800.0
seconds
Quote:
Quote:
Quote:

Equation solver computational rate : 37654.2 Mflops
Equation solver effective I/O rate : 6614.6 MB/sec

Maximum total memory used : 33127.0 MB
Maximum total memory allocated : 36706.0 MB
Maximum total memory available : 48 GB

+------------------ E N D A N S Y S S T A T I S T I C S
-------------------+


I can see that the times in the slow model are higher than those of
the
Quote:
Quote:
Quote:
quick one. Also, the memory usage in the slow is larger than that of
slow
Quote:
Quote:
model. My question is, what controls the memory usage? Can I so
something
Quote:
Quote:
for it?

Your suggestions are highly appreciated.

Thanks,
Mohammad
--
=====================================
Mohammad A Gharaibeh
PhD Candidate,
Mechanical Engineering Department
State University of New York at Binghamton
=====================================
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+




--
Regards,

Saad Aziz
PhD Student
Department of Mechanical Engineering
Louisiana State University, Baton Rouge, LA, USA
Phone: 225-287-0483
E-mail: saziz1@tigers.lsu.edu
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+
+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

+-------------------------------------------------------------+
| XANSYS web - www.xansys.org/forum |
| The Online Community for users of ANSYS, Inc. Software |
| Hosted by PADT - www.padtinc.com |
| Send administrative requests to xansys-mod@tynecomp.co.uk |
+-------------------------------------------------------------+

Post generated using Mail2Forum (http://www.mail2forum.com)
Back to top
View user's profile Send private message
Display posts from previous:   
This forum is locked: you cannot post, reply to, or edit topics.   This topic is locked: you cannot edit posts or make replies.    XANSYS Forum Index -> XANSYS
All times are GMT - 7 Hours
Page 1 of 1

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You cannot download files in this forum


Powered by phpBB © 2001, 2005 phpBB Group

sleek template created by Andrew Charron