Why is Openstudio 1.14 simulation time twice as long as with EnergyPlus ?

asked 2017-01-20 03:25:34 -0600

OS-user-AT

2284 ●3 ●14 https://mak-analytics.com

updated 2017-08-15 13:56:44 -0600

__AmirRoth__

4436 ●5 ●17 http://bleedinggreenna...

I have compared simulation times of an osm-model and of the corresponding OpenStudio export idf-model: The OpenStudio version (osm) took 03:59:23 (hours::minutes:seconds) while the EnergyPlus version (idf) took 01:53:13 for simulating one full year. I expected a slight overhead by running simulations with OpenStudio but this seems excessive and probably unnecessary.

What can developers say to explain this behavior ?
Can this be avoided in future releases ?

UPDATE (due to questions in comments): I took an OSM and ran it through the OS App 1.14 ; then I took that same OSM, exported to IDF, and ran the IDF with EP-launch. The run times to be compared are those given in the respective err files.

edit retag flag offensive close merge delete

Comments

Where did you pull the *.idf from? 'EnergyPlusPreProcess' folder? Or did you do a *.idf export?

Adam Hilton ( 2017-01-20 08:12:18 -0600 )edit

This does seem odd, should be small overhead. I assume you are using EneryPlus 8.6 to run the IDF, and are using the same hard drive that the OSM project is on?

David Goldwasser ( 2017-01-20 09:39:58 -0600 )edit

I used the normal idf-export, not the in.idf (which might be not identical, although they should). And I used EnergyPlus 8.6., of course on the same local harddrive.

OS-user-AT ( 2017-01-20 10:10:58 -0600 )edit

Do you have measures in the measures tab? These won't be reflected using the export function.

Adam Hilton ( 2017-01-20 10:44:24 -0600 )edit

1

how many reporting variables (and what reporting frequency) do you have in the model?

BrianLBall ( 2017-01-20 11:28:37 -0600 )edit

see more comments

3

answered 2017-01-24 14:30:08 -0600

David Goldwasser

20395 ●94 ●14 http://www.openstudio.net/

updated 2017-01-24 14:30:53 -0600

The reason this model is running slowly may be related to the 8.5 to 8.6 issue you identified in this post. It also has a number of very small zones with volumes less than 7 ft^3 that might be creating issues.

But we don't have an explanation for, nor have we been able to reproduce, the slowdown you were seeing between EnergyPlus 8.6 (run through the OpenStudio application) and EnergyPlus 8.6 (run through EP-Launch). I used your OSM file and downloaded the Santa Ana epw that your OSM seemed to refer to. I assume you didn't customize that weather file as it wasn't sent with your OSM.

Here is my runtime in OpenStudio 1.14.0 on mac
`Elapsed Time=01hr 43min 43.07sec'

And here is my runtime in EP-Launch-Lite 8.6 on mac
`Elapsed Time=01hr 50min 3.16sec'

In my case the OS app was actually a little quicker, but I was doing other work on my computer for the EP-Launch-Lite run, so that may have negatively impact the runtime by the 7 minute delta from the OS run. I'll update this if we come up with additional ideas on how to replicate what you see. Might be specific system configurations your your machine?

edit flag offensive delete link

Comments

The Santa Ana epw used was not modified, but the ddy was taken from Los Angeles. Since you could not reproduce the run time increase, I'll have to make further checks and tests (e.g. a diff-view between in.idf and export.idf; potential effects from modified IDD,...). I'll let you know as soon as I find something noteworthy. My operating system was Windows10 64 bit.

OS-user-AT ( 2017-01-25 14:55:56 -0600 )edit

Did you edit the EnergyPlus IDD in one of your installations?

David Goldwasser ( 2017-01-26 20:32:21 -0600 )edit

the EP-Launch-Energyplus-IDD was extended to allow for more PV generators, and for running air-water heatpumps at below zero degrees, but it is unlikely this had any effect on runtime increase. BUT see my comment to macumber's experiment.

OS-user-AT ( 2017-01-27 03:08:18 -0600 )edit

add a comment

2

answered 2017-01-27 05:38:44 -0600

OS-user-AT

2284 ●3 ●14 https://mak-analytics.com

updated 2017-01-27 06:05:36 -0600

The explanation for this is most likely the following: The 32bit code of OpenStudio 1.14 - along with the bundled 32bit EnergyPlus code - runs more slowly than the 64bit code of (independent) EnergyPlus 8.6 under Windows10 (64bit).

The rationale for this explanation is:

Running the same OSM with the OS 2.0 App (64bit) is completed within EP-runtime 01:57:23, while the EP-export-IDF from OS 2.0 has a runtime of 01:55:22 - very much as expected. As macumber showed in his experiment, OS 1.14 and OS 2.0 (64bit) will typically have almost the same runtimes. Hence it must be the 32bit code that creates the remarkable time increase.

edit flag offensive delete link

Comments

That seems plausible, nice detective work!

macumber ( 2017-01-27 09:28:15 -0600 )edit

add a comment

2

answered 2017-01-26 18:11:39 -0600

macumber

11968 ●22 ●21 http://automaticmagic.com/

I tested the following cases on Windows 7 64bit and did not see any difference in the run times:

OS 1.14, run from app - Elapsed Time=02hr 35min 43.61sec
OS 1.14, export idf, run in EP_Launch - Elapsed Time=02hr 35min 14.80sec
OS 2.0,  run from app - Elapsed Time=02hr 35min 48.78sec
OS 2.0, export idf, run in EP_Launch - Elapsed Time=02hr 36min 44.26sec

edit flag offensive delete link

Comments

the most plausible explanation for the time increase on my computer, however, could be, that I am running OpenStudio 1.14 32bit under Windows10 64 bit. I might try to install additionally OS 1.14 64 bit (without Sketchup), if it does not interfere with OpenStudio 1.14 32 bit installed in parallel - and test for runtime again. Or test the osm-model once again with OS 2.0 64 bit.

OS-user-AT ( 2017-01-27 03:12:22 -0600 )edit

add a comment

Why is Openstudio 1.14 simulation time twice as long as with EnergyPlus ?

Comments

3 Answers

Comments

Comments

Comments

Your Answer

Training Workshops

Careers

Question Tools

Stats

Why is Openstudio 1.14 simulation time twice as long as with EnergyPlus ? edit

Comments

3 Answers

Comments

Comments

Comments

Your Answer

Training Workshops

Careers

Question Tools

Stats

Why is Openstudio 1.14 simulation time twice as long as with EnergyPlus ?