Beowulf cluster management and creation using Grid?

Just fell into the hole that is cluster computing. Mainly because I have a few computers left over and got interested in parallell computing.

Would it be possible to tweak/add to Zorin Grid in order to use it to manage and set up a Beowulf type cluster (I haven’t fully explored the Grid yet)? Or rather, would that be totally out of scope for Zorin considering RedHat and a few others do that stuff and it’s perhaps too far from “regular” users?

Basically, set up a grid, push the images to the nodes (which I suppose is no problem with regular Grid) but maybe also have “everything” you need for a cluster packaged and ready to run a, let’s say, max 6-8 node cluster? I’d gladly pay a little extra for that, and more for a stable 16 node cluster :smiley:

EDIT: The OSCAR project seems to be defunct but something along those lines. Also, if one could use GRID and Zorin OS to manage this then unused computers could enter the cluster when available. In principal, any school or small company running a few workstations by day could run a computing cluster / render farm et c during off hours. Potentially sell computing power even??


Thanks for the suggestion @Wolf337!

We currently don’t have any concrete plans to add support for managing Beowulf clusters in Zorin Grid. However, if there’s enough interest, it’s a feature we could add to our roadmap down the line.

Please also note that, as Beowulf clusters run tasks on separate systems without shared memory, the software running across the cluster would need to be written specifically with these limitations in mind. This would only be feasible in specialised use cases.

1 Like

Thank you for that quick answer :slight_smile:
Very glad to hear that it’s not completely off the table.

I assume that since Grid allows (“will allow”, looking forward to the release!) me to manage all computers on my network, that would also be true for any virtual machines I set up?
Simplest solution would perhaps be to use a minimal pre-configured Linux installation on a KVM running on the clients (with a Grid auto-deploy option). The nodes can then run 24/7 invisible to the users and the resource allocation can be done to match a minimum specification across all hardware (if that is deemed beneficial for Beowulf clusters).
The “only” thing that would be added to Grid would be a “Cluster settings”-tab where cluster host file, shared drive and such preferences are set.
I’ll see if I can boil all the software and steps down to a nice light soup and make it as simple as possible :smiley:
Never know, might come in handy. Especially for democratizing availability to “super” computing for AI, FEM, fluids, rendering and other fun stuff.

Live long and prosper! :vulcan_salute:t2: