Virtualization of FlexSim on a Raspberry

[clair_augsburgerZGWRZ]

[ FlexSim 16.2.0 ]

Hi, one engineering school is testing a solution to virtualize FlexSim. Although they allocate several cores, if they launch several parallel instances of FlexSim only 1 core is utilized. Has anyone already tested this kind of configuration of virtualization ?

Here are more details on the setup:

Raspberry: RBPi 2 with Thinlinx

Server: Citrix solution, hypervisor XenServer VM (2 VM per server), Windows server 2008 R2, 8vCPU, 32Go, nVidia Grid K2 passthrough, normally for 16 users simultaneously on the virtual machine

[Ben_WilsonADSK]

I think mentioning Raspberry Pi is too much information and confuses the situation. It would seem to me that the OS of the client initiating a remote session is not important. Raspberry Pi + FlexSim is so strange, it scares people from looking deeply into the question.

If I understand the situation correctly, remote clients are logging in to separate, concurrent Windows Server 2008 R2 sessions, and they are starting FlexSim. As they use FlexSim, someone else is looking at CPU utilization on the server and finding that even though multiple users are logged in using multiple instances of FlexSim, only one CPU core is being used.

Is this correct?

Does Windows Server 2008 R2 give a breakdown for which CPU cores are being used? Or are you simply inferring that 12.5% CPU utilization on your 8 core machine means that only 1 core is being used?

How many users have been observed logged in and running FlexSim simultaneously where CPU utilization never exceeds 12.5%? Are users running models, or just building them? If multiple users are running models simultaneously, CPU utilization still never exceeds 12.5%?

[Ben_WilsonADSK · Accepted solution]

FlexSim's dev team may have more insights than I do, but the fact that none of them have commented on this question suggests they don't have any answers, either.

This answer at superuser.com has some relevant information regarding threads, processes, and CPU cores. Basically, it comes down to the Operating System being in charge of where it allocates threads.

If you look at the task manager on the Windows Server, when several clients are remoting in and using FlexSim, you should see several FlexSim Processes listed. It is the Operating System's duty to allocate these processes and their threads to the hardware.

My understanding of FlexSim's Experimenter and Optimizer is that FlexSim launches as many new FlexSim background instances as there are CPU cores. It is then the Operating System that takes care of handling how the system runs the instances. FlexSim itself is not assigning certain instances to certain CPU cores.

Any further insights, @anthony.johnson, @phil.bobo, @jordan.johnson?

[anthony_johnsonT83CM]

That sounds about right to me.