How can I build a Matlab parallel computing cluster from computers with existing standalong Matlab installations?

Hello, recently I have a large computing problem which I hope to be speeded up by running it on a computer cluster. We don't have an existing cluster so I think I need to build one from scratch. What we do have in our lab are several spare, fast computers, each having a standalong Matlab installation, and of the same version (R2018b). They all have parallel computing toolbox installed. I hope to run my parallel computing jobs on these computers, but does not affect existing Matlab installations on them, as someone else still need to run some light-weight Matlab jobs directly on them. My intention is to run my Matlab job as a background process on these computers using all of their spare CPU time, but these computers does not belong to me so I don't want to affect the usage of their original owners.

However, after reading the documentation of Matlab distributed computing server, I feel a bit confused about how they work. It seems that they need a separate Matlab installation and a separate license on each computing node. What's more, it seems that the computing node cannot run regular Matlab desktop session any more.

How can I build a Matlab parallel computing cluster from computers with existing standalong Matlab installations?

Answer 1

[Jeff Miller]

> after reading the documentation of Matlab distributed computing server, I feel a bit confused about how they work

Yes, me too.

if your parallel computing problem can be broken down into many different function calls that can be run in any order and that will each produce its own output file (which can be summarized after all are done), you might be able to use BudgetCluster . It doesn't use the parallel computing toolbox at all, so requires no special set-up.