I am an architectural design student who works with 3D programs (like Blender, Rhino, etc.). I usually find that it takes a lot of time to compute heavy geometry. So, I thought about if we could distribute the computations to a number of machines to get the job done quickly.
Now the problem is, I wanna distribute the computations without actually running an instance of the software I am using in every machine. Because, I have read that a lot of distributed systems work that way, by running an instance of the software in every node that is doing the job. But, my question is, can we run the 'master software' in only one machine, and just use the resources of many machines by running a 'supporting software' (that we write) on every machine to get the job done. So its like treating 'n' number of machines as 'one'.
Is it possible...?
Don't think of them in terms of machines, but in terms of resources. So instead of needing to run software on a machine, you need your software to be able to make use of a resource from a machine within your software's reach (say the network your "server" resides on) instead of the actual machine it's running on.
Unfortunately, most workstations do not expose their graphics card as a service. Unless you implement something like vGPUs (https://www.nvidia.com/en-us/design-visualization/technologies/virtual-gpu/) I don't see what you need happening.