It’s process based, so communication overhead is heavy and the infrastructure is minimal - if you’re distributing jobs which are individually large it works fin

It’s process based, so communication overhead is heavy and the infrastructure is minimal - if you’re distributing jobs which are individually large it works fine, but it can’t really be used to replace multithreading in a shared memory model.