Jump to content
Simio Forum

bmarlin

Members
  • Content count

    23
  • Joined

  • Last visited

Community Reputation

0 Neutral

Converted

  1. I have been running very large experiments that take a few days to run. Murphy's law seems to interfere through either power outages or automatic updates which can cost me a couple of days of computer time. So I found some code using autohotkey and modified it to work great with SIMIO. autohotkey is a free download and easy to use. The below code runs a loop which hits {control S} on the SIMIO window if it is open every 10 minutes. Persistent ;run a loop 100 times saving every ten minutes if your SIMIO file is open. loop, 100 { ;save file every 10 minutes MsgBox, Iteration number is %A_Index%. setTimer, AutoSave, 6000000,on AutoSave: IfWinActive, ;name_of_simio_file_without_spfx_extension Send {Blind}^s return } return
  2. I would like to repeat particular replications in a large experiment. For example, I am conducting 7 replications of 47 scenarios and replication7 of scenario3 is providing some rather disturbing results. Is there any way I can just re-run this particular replication with its seemingly very unique random number seed - maybe even in the facility view, so I can watch this behavior. Thank you.
  3. Because others have looked at this, I figured I could post my work around. I just wrote a quick little script in AutoHotKey which autosaves the experiment at time intervals, so I do not lose all my work when my experiments run over night. I also used R to modify the very large CSV files that the Pivot grid outputs (removes all the NaNs and O values that I do not need). In all this was a couple hours of work and about 9 lines of code. The fact that SIMIO can be used with so many other tools without issue is great.
  4. Hello, My experiments take a rather long time to run (up to a few days). Is there a way to have the Pivot Grid automatically write to a file while the experiment is still running, for example when a scenario is complete, or even at a time interval. I know I can get some of the same functionality through a Write Step, however I have built my output so that it formats well to the Pivot grid. If I can write to a share drive I (1) prevent loss of data due to hardware or software issues and (2) I can do analysis as data is released at scenario completion - even while my experiment is on going. Thank you. Ben Marlin
  5. I have a slightly similar problem which I would like to add to this. I have a decent sized data table of distances that I have built in Excel - it is 400 X 400. My issue is SIMIO seems to have difficulty binding the table. Of course my goal is to be able to make changes to the Excel table as needed, but right now I have waited over an hour and SIMIO seems to still be working to bind the table. Has anybody had a similar experience? Thanks
  6. bmarlin

    Replication Runner and Experiments

    I sent the model to the SIMIO support email address yesterday. Ben
  7. bmarlin

    Replication Runner and Experiments

    [quote]Do you happen to know how many hours pass between starting the run and seeing these problems?[/quote] Yesterday I let the model run for 8 hours and then I left - when I came in this morning same error message. I tried running it on external and 32 bit external. I do not believe the model even completed the Initiation - as I have over 20K entities and a warning message used to be displayed when I ran in experiment mode.
  8. bmarlin

    Replication Runner and Experiments

    Thank you - very informative link. I still have one question. I often get this error [b]No Computation nodes available to run replication[/b] when I apply the advaced settings to: Concurrent Replication Limit:0 Replication Runner:External Distribute Runs: False I am using a 64bit Machine with 8 GB RAM and an i7, and I want to keep the jobs on this computer and let SIMIO choose the number of concurrent runs. The model is pretty big and when I run just 1 replication it can take 6 hours. I just need to know what the error means - hoping it is an easy fix. Thanks, Ben
  9. bmarlin

    Dynamic Combiner Parent Selection

    Thank you! That was the problem! Awesome. Ben
  10. bmarlin

    Dynamic Combiner Parent Selection

    Hello, The gist of this problem is assigning teachers to classrooms based on a student teacher ratio using combiners to bring students and teachers together. The problem is if not every student is batched, he or she is left behind...I would like to prevent or at least minimize this. I have 12 combiners (grade levels) who each receive members from separate servers (the previous year). All parents (teachers) are released from one node at the same time (beginning of the school year) which is a set time - established via a monitor. My problem is I want to evenly distribute the parents across all combiners with a ratio of members to parents - each combiner has a different quantity of members. I have no problem with the dynamic ratio as the required number of members per batch and I have gotten pretty close to a correct answer with the attached screen shot, but it is not 100%. For what ever reason, my model will sometimes ignore any rules and assign a parent to a combiner which does not really need it. I have also tried to use a process which searches the combiner member input buffers and uses a set node and transfer step to assign the correct parent - but without luck. Any ideas would be greatly appreciated. This is close to the One Queue For Multiple Servers simbit, but you cannot make the input buffer 0 for a combiner.
  11. bmarlin

    Replication Runner and Experiments

    I am looking for a bit more clarity regarding running experiments and maximizing the use of a 64 bit machine. Right now I am running on a i7 machine with 8 gb of RAM. Using mid level models when I conduct one replication I see substantial run time improvements when I set the advanced options to: 0, External, False. But I have a couple of questions. 1. With a larger model I have been getting an error: No Computation nodes available to run replication. This is after the model has run for hours and I see the replication runner using over 5GB of RAM on my computer. 2. I have seen when I require multiple replications, often the second and third replication take longer than the first. I thought they would all finish at around the same time as the jobs should be farmed out in parallel. Is this a correct assumption? I am still working on gaining permissions to run the replication runner on multiple machines - but I thought it crucial to get established on one machine first. Thanks for any input.
  12. bmarlin

    Experiment vs Facility Runs

    I am using an older version of SIMIO - the support folks were very helpful and said this is probably the reason for the error. However, I now understand the convert to repeat group capability - very cool. I will update this post once our IT updates the model version. Thanks!
  13. bmarlin

    Experiment vs Facility Runs

    Thank you, I have sent the file into SIMIO Help. In the meantime, I am working to convert lower level tables to repeat groups. The problem seems to lay in a data table which I read in as a matrix. I later reference this array as a distance map - ie I have an understanding of where I am, and I know where an event occurred, so I therefore can look the distance up on a data table. This seems like a property/data table that one would want at the entity level as it is meant as "knowledge". When I make this change, I now get an error because one cannot reference a repeat group property to populate an array - is there a workaround, or do I need to populate this with a set of processes? Thank you.
  14. bmarlin

    Experiment vs Facility Runs

    I have set an experiment with one replication and no control variables - the goal being to see how much faster the model runs in an experiment rather than in the facility view. The facility view model runs fine, however the experiment runs are returning a divide by zero error. This error is pointing in a table which is in a lower level hierarchy object. Is this the cause of the problem? Do all tables need to be at the top level model? Thank you.
  15. OK, I solved this myself. The SIMBIT is helpful and I just needed to wrap my head around the Parent Member relationship. Using this command: [ParentObjectType].BatchMembers.FirstItem.[MemberObjectType].VariableToPass (Without the brackets). Putting this in a process that is called when the batch exits the combiner allows me to pass any information the first item in the batch has with the parent. Ben Marlin
×