Coderjoe wrote:Do you know? Also, how many downloaders do you expect per torrent, on average?
I do know, however I wouldn't expect anyone to take my word. It would be more worth while to research an answer than to be given one that you would never honestly believe. For the mean average about 150 downloads evenly spread over 30 days.
Bittorrent really isn't wothwhile for the mass of videos being downloaded a few times a month. Especially when many people are likely to close the torrent once their download is complete. Additionally, who do you think would be seeding the torrents? More than likely the same download server that is storing the video and currently sends it out via http already.
Those that close without providing a full return should have their acounts limited, to upload only or download equal to upload. Management for anti leech control is highly effective. Members that can't upload should be able to donate money for download credits or an upgrade to an unlimited acount. Although even the tighest ISP could handle a few megs uploaded over a month. Or maybe your acount will have download bandwidth added ever 15 days or something for free just for being a member. The seeding would be handled in part by the current server and in part by people downloading. As their is a steady flow of downloads, their will always be someone downloading a file that would be able to upload to others.
Bittorrent might be worthwhile for videos with a large number of downloads per month, or for top bandwidth consumers in a month, but not for the peek on Phade's curve.
Think of it like this: if you already serve all that content each month, by changing the way it's downloaded from HTML transfer to BT, what is lost?
After talking this problem over, the solution offered was to convert the whole database into a single superset torrent, and provide single file access torrents (IE one video) for download. People would be able to grab single files and might in passing upload to others, (for sure if they actively choise to help) while the very popular files would mostly transfer themself.
The seeding of the whole database would be covered by the current http servers while the tracking might need to be offloaded onto another system or one of the freed transfer servers.
Try closer to 30TB
30TB in 30 days. I could offer a link to a tracker that passed 70TB in a day.
I doubt there is much of a technical issue with tracking that many torrents. Now, let's talk about seeding even 30,000 torrents. I'm pretty confident the disk arrays required to support random access to 2TiB at 30 MB/sec are confortably out of our price range.
And yet somehow this site currently runs... Further the most common vidoes would not require any access help.
However this is speculation, Phade having taken the effort to dig up a graph and post it, Combinded with his own bad BT experience will likely mean that AMV dot org will never have any BT support, ever. That's my own experience on human nature talking here so I might be wrong.
My parting suggestion is that someone finds what are the top 100 videos by bandwidth, and just go ahead and set up a tracker and make the torrents for them. Ask nicely and have a BT section added in the download options and take at least some of the load off.
If even a fraction of 1% of the people here are willing to offer money, I know that many more would offer bandwidth.
