I'm trying to understand what it means to my viewers where the Wowza server they connect to is located. In the examples below I know I should account for some overhead, but let's keep that out to leave the math simple. In all cases a connection means a connection to a 1000 Kbps RTMP stream.
If I set up a Wowza server at:
- ... home with my 50 Mbps upload, I can theoretically serve 50 connections.
- ... work with our 300 Mbps upload, I would theoretically be able to serve 300 connections.
- ... a local datacenter with dedicated 10 Gbps upload, I would theoretically be able to serve 10,240 connections.
- ... Amazon on a single EC2 instance with dedicated 10 Gbps upload, I would theoretically be able to serve 10,240 connections.
- ... Amazon, and then distributed the stream to edges on another 100 EC2 instances with dedicated 10 Gbps upload, I would theoretically be able to serve 1,034,240 connections (101 servers each handling 10240 connections).
- ... Home, work, local datacenter or Amazon on an EC2 instance, and then distributed the stream to Akamai (or another live streaming capable CDN), how many connections would I theoretically be able to serve?
Now, one thing is how many connections I can theoretically handle, another thing is how many connections will actually be possible in a real world scenario. With regards to something as network congestion, how would a Wowza server at home, work, the local datacenter, an EC2 instance, a fleet of EC2 instances compare? And how would such setups perform compared to using Akamai's CDN?