Production is complicated and ends up being what you want to define. We use a lot of tricks in our infrastructure at the production level to address load balancing of services, etc. Regarding what you would need to run it in a local environment, the default ports are:
8001 - Companion app HTTP server
8004 - aiStarter HTTP server
8888 - App Engine dev server (uploading to App Engine will run it on port 80)
9990 - Buildserver
11211 - memcache (only needed for rendezvous server, not accessed from elsewhere)