Jump to content

How to set server quantity after set work schedule in experiment


Sean

Recommended Posts

Hi all, I met some questions in building the model. It seems a very general quesiton but I didn't find any solution from internet... 

I'm building a general workflow model. I want to decide the quantity of a server to optimize the workflow.

I want to use experiments or optquest to get the 'best' number for each server. If I set the capacity type as Fixed, then I can set the initial capacity as ''referenced property'', then I can use this as control in experiments/optquest.

However, I have a work schedule for each server now which means I couldn't set the capacity type as Fixed. And then I couldn't set capacity as ''referenced property''. I don't know how to control this capacity in my experiments/optquest. 

I know I can change the 'value'(which means the capacity) in workschedule table. But I can not change this number flexible in experiments... Or I don't know how to set this value as ''referenced property'' . 

Is there anyone who could help me with this?  Thanks a lot!

  • Thanks 2
Link to comment
Share on other sites

11 hours ago, jzhou said:

Dear Sean, you can create a new referenced property on WorkSchedule property of Server. and in Experiment windows reference to various pattern-based workschdule name.

image.png.994be2729dea264d7684680898e6cd63.png

image.png.92cd2ce60bba022d5f52a753f27dbe99.png

Best regards/Jeff

 

Thank you very much, Jeff.

So you are saying that I should make up many different schedules in the schedule table first, then running experiments, I can choose different ones to test, right? That helps and definitely would work for me. But it's more manual, not like in OptQuest, I can just set lower bound and upper bound to kinda automatically run the experiments... Do you know any ways to do that?

Even there's not, your method is still great for me! Thank you again!

Link to comment
Share on other sites

To the best of my knowledge, optquest has no capability of optimizing string or boolen type controls. In the case of automatic optimization (which you want) it does not work! (i.e., It works automatically with only numerical values.). I do not understand why it can not but this is the way it optimizes so far. If you want to optimize string type controls you have to code your optimization routines (i.e., usually via metaheuristics) and integrate them with simio. For this to happen you should be somewhat an advanced user. If not you should ask for further help. 

Link to comment
Share on other sites

On 2/5/2019 at 10:31 PM, Sean said:

Thank you very much, Jeff.

So you are saying that I should make up many different schedules in the schedule table first, then running experiments, I can choose different ones to test, right? That helps and definitely would work for me. But it's more manual, not like in OptQuest, I can just set lower bound and upper bound to kinda automatically run the experiments... Do you know any ways to do that?

Even there's not, your method is still great for me! Thank you again!

Maybe you could create several property of type 'Interger' and reference them in the value row..like below... and in the experiments you should be able to see them as controls to optimizer.

image.png.0ebecfa1b218bf790e0e5022eaaffb13.png

Best regards/Jeff Zhou

 

  • Thanks 1
Link to comment
Share on other sites

×
×
  • Create New...