Virtualizing Matlab under Microsoft AppV

We are trying to virtualize Matlab using App-V by Microsoft. The sequencing goes fine, the newest App-V 5.1 packages also the Vcc runtimes and Java is not a problem when loading the application. The only problem we found is that packaging the whole Matlab+toolboxes takes ages to start. Anybody did successfully sequence it ? thanks. alessio

[Ian Sadkovich]

We are having similar issues. Matlab 2014 wasn't exactly speedy with App-V 5 but with version 2015 its an absolute dog.

Mathworks, before you say that something is not supported, please do your research about the technology. App-V isn't a virtual machine.

[Mark Hodkin]

This doesn't take away from the support issue. That is a different subject........ I found if I sequenced (App-V 5.1) with the PVAD enabled and set to C:\Program Files\MATLAB\R2016a then the launch times is rapidly reduced. This may be still too slow....but a lot faster

Without PVAD set 31 minutes 25 each launch With PVAD set 4 minutes 22 First launch Second launch 58 Secs Physical Install 42 secs

I also Stream faulted the application rather than fully download or setting FB1 improved the performance.

To set the PVAD in App-V 5.1 From a command prompt run Sequencer.exe -EnablePVADControl

As mentioned in other posts the number of toolboxes included also seems to have an impact.

For Timings and debugging I used Tim Mangan's AppV Manage tool (<http://www.tmurgent.com/Tools/AppV_Manage/Default.aspx>) #

There are nearly 300 fonts that are included in Matlab so I may also experiment with installing them locally as I've seen performance increases with other applications when the fonts are installed locally. Other avenues to follow up also include ensuring all the Visual C++ packages are installed locally so no additional side by side work needs to take place.

Hope this helps

Answer 1

[Rahul Goel]

Hi Alessio,

Virtualizing MATLAB using Microsoft AppV is not a supported workflow by MathWorks. Also, from the steps you mentioned, loading multiple toolboxes with MATLAB is a memory intensive task and can be a long process on a virtual machine.

[Alessio]

Dear Rahul, thanks for your replay. It's a pity that Mathworks does not support such installations since nowadays it's becoming more frequent to find virtualized environments. Anyhow, I found out that starting Matlab from the MatlabAppStarter takes much more time that launching Matlab from the binary inside the "win64" folder. Do you maybe know if that is needed? What is doing the MatlabStartApplication?

Thanks, alessio