Back in the day, there was a distributed cluster OS called Mosix. Even back then I had several spare computers lying about, and the idea of being able to chain them all together and have one virtual computer that would automatically distribute processing without special coding was enticing. It turned out to not work very well unless you did specially code for it, or clustered the computers very tightly with fiber; it just wasn’t worth it.
But when I see piles of compute like this, a part of my still wants to network them all together and run … well, whatever fills the shoes of OpenMosix these days, if anything does.
Back in the day, there was a distributed cluster OS called Mosix. Even back then I had several spare computers lying about, and the idea of being able to chain them all together and have one virtual computer that would automatically distribute processing without special coding was enticing. It turned out to not work very well unless you did specially code for it, or clustered the computers very tightly with fiber; it just wasn’t worth it.
But when I see piles of compute like this, a part of my still wants to network them all together and run … well, whatever fills the shoes of OpenMosix these days, if anything does.
Some modern workloads can take advantage of multiple computers. You can usually compile using things like distcc and spread the load across them.
If you make them into a Kubernetes cluster you can run many copies or many different things.
It’s still an unsolved problem: we still end up with single core bottlenecks to this day, before even involving other machines altogether.