On Mon, 12 Feb 2007, Paul Vixie wrote:
I never quite understood why layered multicast never took off which would solved the problems you state above. There have been so many research papers on the subject from the late 90s that I would have thought that by now IPmc would be the silver bullet for video distribution.
as i said earlier, for intranet use, ip multicast is all the rage for video content. i'm fairly sure it was in use at my hotel in cairo last week, and i know it's been deployed in a number of "digital television" networks in asia. it's internet multicast (idmr) that never happened, and as far as i can tell, that's because there's no billing or business model for it.
Why couldn't internet multicast be used for content other than video? Stream Torrents, .mp4 files, etc. Instead of just sending a single video stream at some data rate, stream data files sequentially. Stream owners can post a schedule (or not, just sending a stream of files with metadata headers), your pc-based "TiVo-like" software can tune in (request the stream from your provider, which turns on and off all the streams they receive and only sends requested streams to your "Last Mile" on request) based on that schedule or request. NBC can now stream their shows to me as a .mp4 and I could grab them as fast as they could send it, rather than in realtime. They might offer the same stream at different data rates: 1mbps, 5mbps, 10mbps, 30mbps (for those of us lucky enough to have Verizon FIOS at home). The streams would simply repeat once they streamed all the files in a list. Think of a YouTube stream. As videos are uploaded, they are encoded and sent out an internet multicast stream. It's not a video stream, but a file stream, where one file is sent right after the other, and your end receiver knows what to do with the data. Metadata is put into the file headers so you can scan for content/description. Your "TiVo" can pickup the videos you might like to watch based on your keywords, and now you can watch those videos on your TV on demand, already on your PC. YouTube only had to broadcast it once, and thousands of people who may get the YouTube stream have decided to keep it or not. Sure, it might take up lots of disk space, and analyzing a stream (or 10 simultaneously) might take up a bunch of CPU/memory, but it'd be a way to distribute content efficiently and potentially lower transit bandwidth usage as people started to use it rather than today's status quo. If a channel is popular enough, people ask their provider to carry it. The provider is incentivized to carry a channel if the bandwidth they utilize to serve the unicast version of that data is greater than the amount of data they might use for a single multicasted stream of that same data. Rather than the end user paying for it, the provider saves money by utilizing the stream. Beckman --------------------------------------------------------------------------- Peter Beckman Internet Guy beckman@purplecow.com http://www.purplecow.com/ ---------------------------------------------------------------------------