sensitivity analysis (vensim)

Use this forum to post Vensim related questions.
RWTH_FCN
Senior Member
Posts: 148
Joined: Wed Feb 08, 2017 9:10 am
Vensim version: DSS

Re: sensitivity analysis (vensim)

Post by RWTH_FCN »

I ran the model on a super computer but since this version does not do any parallel multicore computation, the running time did not reduce. Do you have any plan for releasing that experimental version?
tomfid
Administrator
Posts: 3804
Joined: Wed May 24, 2006 4:54 am

Re: sensitivity analysis (vensim)

Post by tomfid »

We're working on it, but I'm not sure whether it'll be done by end of year. The speedup is more than noticeable if you have a many-core machine. Older workstations with 16 or more cores are now readily available at dirt cheap prices. We'll keep you posted as this evolves.
RWTH_FCN
Senior Member
Posts: 148
Joined: Wed Feb 08, 2017 9:10 am
Vensim version: DSS

Re: sensitivity analysis (vensim)

Post by RWTH_FCN »

tomfid wrote: Mon Oct 14, 2019 5:24 pm We're working on it, but I'm not sure whether it'll be done by end of year. The speedup is more than noticeable if you have a many-core machine. Older workstations with 16 or more cores are now readily available at dirt cheap prices. We'll keep you posted as this evolves.
Thanks for your information.
RWTH_FCN
Senior Member
Posts: 148
Joined: Wed Feb 08, 2017 9:10 am
Vensim version: DSS

Re: sensitivity analysis (vensim)

Post by RWTH_FCN »

Hi,

I did Monte carlo Simulation for a large-scale model, after many simplifications in my model I got this error (it happens at Sensitivity Simulation 1/3):

error writing results-disk may be full

I was wondering is there any solution for it or is it related to size of model?
tomfid
Administrator
Posts: 3804
Joined: Wed May 24, 2006 4:54 am

Re: sensitivity analysis (vensim)

Post by tomfid »

Any chance that the disk really is full (or would be, after writing the results)?
RWTH_FCN
Senior Member
Posts: 148
Joined: Wed Feb 08, 2017 9:10 am
Vensim version: DSS

Re: sensitivity analysis (vensim)

Post by RWTH_FCN »

tomfid wrote: Mon Oct 28, 2019 5:44 pm Any chance that the disk really is full (or would be, after writing the results)?
I don't think so. Maybe it is related to the size of model
tomfid
Administrator
Posts: 3804
Joined: Wed May 24, 2006 4:54 am

Re: sensitivity analysis (vensim)

Post by tomfid »

Are you using 32 or 64 bit Vensim? If you do a run with only 1 sensitivity instance (if you can), how big is the file size increment over a non-sensitivity vdf file?
tomfid
Administrator
Posts: 3804
Joined: Wed May 24, 2006 4:54 am

Re: sensitivity analysis (vensim)

Post by tomfid »

Actually, I see that the error happens on sim 1. Looking at the code, the counter for writing sensitivity result blocks is an integer, which will impose a 2 gigabyte limit per simulation. Any chance you're near that?
RWTH_FCN
Senior Member
Posts: 148
Joined: Wed Feb 08, 2017 9:10 am
Vensim version: DSS

Re: sensitivity analysis (vensim)

Post by RWTH_FCN »

tomfid wrote: Tue Oct 29, 2019 7:09 pm Are you using 32 or 64 bit Vensim? If you do a run with only 1 sensitivity instance (if you can), how big is the file size increment over a non-sensitivity vdf file?
I am using 32 bit now. The problem occures when I increase number of iterations. I think your assumption is correct, when size of vdf reaches 2 Gb it produces that error.
tomfid
Administrator
Posts: 3804
Joined: Wed May 24, 2006 4:54 am

Re: sensitivity analysis (vensim)

Post by tomfid »

Can you increase the SAVEPER or thin out the savelist a bit? If you're getting gigabytes of output from a one or two simulations, it'll be pretty unwieldy when there are many.
RWTH_FCN
Senior Member
Posts: 148
Joined: Wed Feb 08, 2017 9:10 am
Vensim version: DSS

Re: sensitivity analysis (vensim)

Post by RWTH_FCN »

tomfid wrote: Tue Nov 05, 2019 3:50 pm Can you increase the SAVEPER or thin out the savelist a bit? If you're getting gigabytes of output from a one or two simulations, it'll be pretty unwieldy when there are many.
Sorry for late response. Actually, I used increased version of SAVEPER. For my case, as long as the model is huge (simulation of compiled version takes around 20 minutes) implementing Monte carlo and stochastic optimization is a big deal in Vensim.
Post Reply