Netflix's recent push to encourage streaming only video plans has brought up concerns that ISPs may not be able to provide adequate bandwidth. This made me recall a post I did on my private blog some years ago about reusing our broadcast infrastructure to provide personalized video viewing (or at least the bulk of it). It is loosely modeled on pay per view techniques, but assumes every subscriber has a DVR (digital video recorder) and eliminates channels as we know them (except perhaps for real time things such as news) and just considers the stations as data streams. Users select which shows and movies they would like and then a central program would determine the best way to send the shows over the available stations so that the viewers' DVRs could record them, accounting for collisions. As an option, viewers could pay more to get a higher priority in the queue if their content is more unusual. Optionally, a recommendation system could be employed to gather content when free capacity is available (it would be cleared from the DVR's drive automatically if space was later needed). Commercials should be eliminated from program downloads, but could be spliced in afterwards, allowing them to be varied or omitted altogether if a higher price is paid. This requires very little new infrastructure and if direct data is delivered, the hardware could be cheaper than traditional DVRs, by eliminating the video capture elements as some satellite boxes already do. Adding a secondary tuner is also common, and would allow for more flexibility with regard to scheduling. Of course, once all the content is digital, many other optimizations are possible, such as interleaving of programs, or more frequent broadcasts of the beginning of a show (enabling the rest to be downloaded in the background as the first part is being viewed). Only a very low speed connection would be needed to upload show requests, so dial up would be adequate for areas without high speed internet. Another advantage of this approach is that all times are equally valued, as the content is inherently time shifted and a lot of uncertainty about show rating measurement would be removed. Real time programs such as sports or news could in theory work the same way, but would be biased for earlier delivery since such content loses value quickly. This system is close to a true individual content distribution method, but sacrifices instant viewing in order to exploit redundancies, reducing overall bandwidth requirements.
Monday, November 29, 2010
Broadcast Based Personalized Video Distribution System
Netflix's recent push to encourage streaming only video plans has brought up concerns that ISPs may not be able to provide adequate bandwidth. This made me recall a post I did on my private blog some years ago about reusing our broadcast infrastructure to provide personalized video viewing (or at least the bulk of it). It is loosely modeled on pay per view techniques, but assumes every subscriber has a DVR (digital video recorder) and eliminates channels as we know them (except perhaps for real time things such as news) and just considers the stations as data streams. Users select which shows and movies they would like and then a central program would determine the best way to send the shows over the available stations so that the viewers' DVRs could record them, accounting for collisions. As an option, viewers could pay more to get a higher priority in the queue if their content is more unusual. Optionally, a recommendation system could be employed to gather content when free capacity is available (it would be cleared from the DVR's drive automatically if space was later needed). Commercials should be eliminated from program downloads, but could be spliced in afterwards, allowing them to be varied or omitted altogether if a higher price is paid. This requires very little new infrastructure and if direct data is delivered, the hardware could be cheaper than traditional DVRs, by eliminating the video capture elements as some satellite boxes already do. Adding a secondary tuner is also common, and would allow for more flexibility with regard to scheduling. Of course, once all the content is digital, many other optimizations are possible, such as interleaving of programs, or more frequent broadcasts of the beginning of a show (enabling the rest to be downloaded in the background as the first part is being viewed). Only a very low speed connection would be needed to upload show requests, so dial up would be adequate for areas without high speed internet. Another advantage of this approach is that all times are equally valued, as the content is inherently time shifted and a lot of uncertainty about show rating measurement would be removed. Real time programs such as sports or news could in theory work the same way, but would be biased for earlier delivery since such content loses value quickly. This system is close to a true individual content distribution method, but sacrifices instant viewing in order to exploit redundancies, reducing overall bandwidth requirements.
No comments:
Post a Comment