It is not a grid. It is just a process supervisor which can be controlled from a central point. Nothing more, but it does the work well. You can of course integrate it into a larger system, like what I have done for a PaaS:
Thanks for that link, it was very helpful. It got me wondering if one could run most of your PAAS stack on dedicated hardware at a colo, and the Mongrel2 frontend(s) would sit in a small EC2 instance(s). Could give me your opinion on such a setup?
You can definitely do that as you will get inter node communication over the EC2 private IP. You can define your EC2 firewall to also segment your production/staging/testing setup. This would be very elegant.
Question of taste I suppose. In this case, I needed a simple way to manage a lot of dynamic processes. Normally you have process supervisors which are tuned to run your MySQL, MongoDB or whatever. That is, you would create one profile, often a file or a collection of file per process. With diprocd a given process is just a dictionary in a JSON file which allow ease of management of 100's of processes which change on a regular basis (each push on the PaaS creates a new series of processes on the nodes).
