I've been working on the algorithms for scaling up Cinemon to work for large systems (half a million subscribers or so) on and off. Today I decided to work on the overall architecture, rather than the nitty-gritty details of how communications will work among the federated servers.
The actual "how" of it is something I'll have to keep quiet about, but suffice it that I realised today I was trying to do entirely too much. People watching 500,000 modems (maybe 1,000,000 subscribers) don't need or want that one server tracking every field of every modem.
There's two major strategies for running, one assumes good (very fast) communications among the federated servers, the other assumes poor communications with extremely large resources on the presidential server. I'm guessing that the first case will be the more common and the second less so, but I'm planning on building both approaches and allowing the system to blend between them as required.
Anyway, I think I've pretty much reached exhaustion point, not to mention hunger point, so I guess that will be the day.
Pingbacks are closed.