What about Signal Acquisition?
In simple terms, signal acquisition into Amazon EC2 has to be facilitated on IP. Initially we used both videoconferencing and broadcast MPEG transport stream video sources to contribute signals into the infrastructures directly from events. Likewise we used voice over IP (VoIP) for audio-only signal acquisition.
Today, we have delivered a “field encoder” image - our Out-Side Kit (OSK) - that combines a simple web GUI on top of a minimum configurable encoding engine to avoid engineering issues in the field. The cloud-core system configures each remote OSK for each job. The OSK uses GRIT (our multipath channel-bonding protocol) to aggregate available IP into a single logical pipe, and can encode from low bitrate audio to 4 k. Each OSK contributes to multiple points in the Amazon network, and we also have a virtual telephony bridge that acquires a backup audio stream from the “normal” analysts' conference bridge in case the IP fails all together. This can be muxed into the video streams or simply delivered as its own audio stream to the web users.
The Amazon web services data centers are all extremely well connected, so contribution issues have usually been down to remote connectivity at the origin (on-site at the event). We have, at times, experienced network splits within the Amazon infrastructure, emphasizing the importance of being able to automatically select alternative zones for encoders on-the-fly.