Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Adobe Flash To Eliminate Bandwidth Costs With P2P (torrentfreak.com)
38 points by ukdm on May 19, 2010 | hide | past | favorite | 35 comments


I'm not a huge fan of flash, but I do think peer-to-peer is a great way for the web to grow. Anything legitimate - making it harder for ISPs to say "it's only for pirates" and throttle it - is welcome in my book.


Actually, as someone who used to help run a mico-ISP, the problem with P2P for ISP's has little to do with piracy.

The major problem is it can absolutely rape networks - and it only takes one or two abusers to clog it up for everyone else.

In the end we simply kicked anyone who consistently ate up bandwidth with P2P - and network reliability improved no end.


Help me understand this. Did you not have bandwidth caps?

I understand how 'one or two abusers' could clog up a network at, say, a university where there might not be per-port or per-device limits on bandwidth.

But I pay my $40/mo. for 10 down 1 up, and I never get more than that. How does, say, seeding WoW updates 'clog it up for everyone else'?


P2P is a deliberately greedy process - you are consistently opening a large number of connections (as opposed to a single high volume connection).

One of the big issues with P2P is not the downloads per-se but the discovery process - that can seriously hammer the network (imagine how you stress test a server!)

Also; it's entirely infeasible to offer a service as you describe - partly because networks just don't work like that (this answers the other commenter too).

ADSL is a shared line; what other people are doing can adversely impact you regardless of the networks theoretical capability.


As someone who has studied P2P, none of this make sense to me. If I remember correctly, the Bittorrent protocol tries to maintain only a handful of connections at any given time. Parallel browser downloaders are capable of opening more connections.

Also, I don't see how the discovery process would stress a network. I would expect the actual download to place much more load on the server. For instance, I could be wrong, but I think a bittorrent client only needs to contact a tracker to discover other peers. Every iteration after, the client will try exactly one new peer. That hardly sounds stressful at all!

The last point you make is the only one I agree with. You've offered your users more theoretical bandwidth than you can actually deliver, which is common practice. Whether they are trying to use that bandwidth watching youtube videos or downloading with a P2P client, they will be using more bandwidth than you can handle.


Yeh, I probably didnt explain myself well..

> Parallel browser downloaders are capable of opening more connections

This is mostly what I was referring to. Light P2P users are not a problem - it's the heavy users that cause impact. We had a guy who was torrenting 1000's of files; essentially flooding the network with connections.


Oh, well, that's a different problem entirely :). Although I guess it's much harder to start 1000 downloads in a browser, theoretically it's possible!


That makes no sense. If you sold a service, you should have the capacity to provide it.


That actually make sense. What you say is like suggesting that, back in the modem era, an ISP should keep matching number of modems on it's end equal to the number of users it has.

If there's 10.000 users, there should be 10.000 modems to dial into? Business can't survive like this.


I'm not saying there needs to be a 1:1 relation. I'm saying that, if 40% of your users are connected on peak, you should have some number greater than 40% to supply them. (Obviously, you can't account for abnormal spikes in usage). In a normal day, I shouldn't get a busy tone!

In the same way, if you sell me an "unlimited" service, it had better be unlimited. I should be able to saturate my connection (with legitimate traffic; malicious users can (should?) be nuked from orbit) without being disconnected for "excessive" usage. Otherwise, It's not unlimited.


Actually AOL lost a massive class action suit because of its busy signals when they moved from an hourly rate to a flat monthly rate, and then couldn't deliver.

Does Google really have 7.5 GB set aside for every mailbox that exists? No. But people aren't being capped at 5 GB, either.


Deep packet inspection is already way ahead of simple protocol blocking.


But, of course, it's easy to make your P2P packets look like anything you want, so deep packet inspection is just a way to waste electricity.


I wonder how many people will just "click yes" and end up not being able to watch video anymore. In the US of A, there are quite a few connections that are too slow on the upload side to handle P2P while streaming video.

The idea is worthy, but it needs masterful implementation. It takes a lot to balance the upload/download speeds. I am thankful it will be a opt-in. I just hope they throughly explain it to users before they click "yes."



Stratus is non commercial. If you want to use p2p for a commercial project (show ads/charge for something), you need to pay adobe based on the minutes of usage (Seriously).

Its called adobe lifecycle afaicr.


Omegle and chatroulette, both use Adobe Stratus . Both sites depend on advertising revenue.

I dont think Adobe actively bans commercial projects/products from using Straus. Stratus is in beta, and Adobe doesnt provide any official support. So if you want to run a business that uses RTMFP with Adobe support, then LCCS (LiveCycle Collaboration Services) is the way to go.


And Flash Media Server 4 will have RTMFP when it is released. red5 should offer support soon after that.


If this delivers on its promise, couldn't it preserve Flash as a standard for video delivery online?

Cutting bandwidth costs for video websites could be a significant incentive to slow down HTML5 adoption and keep Flash in the game a little while longer.


It all really depends on the implementation.

None of the big players will want tech that doesn't give them enough control over minimum bandwidth/streaming performance.

For small players, putting some h264/ogg files up on Amazon cloudfront may cost some money, but it's extremely easy to set up right and offers extremely good and consistent performance for the money. P2P doesn't add much for distributors that don't get massive amounts of viewers.

Not saying it can't be a game changer somehow; a similar P2P strategy worked out pretty well for Skype. But given that it's all talk right now, I don't think it'll do much to slow down HTML5 adoption for a while.


Unless HTML5 comes up with a UDP websocket protocol :D

That could be fun, but also possibly a huge security concern.


> Unless HTML5 comes up with a UDP websocket protocol :D

And passive mode for WebSocket. And NAT traversal, and a central coordinate server, and fuck it, let's just use Flash.


Given Flash's record, isn't their implementation already a security concern?


The beta of this technology is what powers chatroulette, right?


Chatroulette uses Stratus and RTMP. RTMP, when Stratus fails because of firewall restrictions.


Probably not. To enable this P2P, the user must accept a separate dialog window asking their permission.


chatroulette uses stratus, or at least used it. When it became famous, it completely killed stratus server which was still in beta and not ready for massive usage. Everyone who happens to use stratus at this time complained.

So stratus is using p2p, but it still rely on Adobe servers and as far as I know there is no way to deploy your own stratus server yet.


It's been coming for over a year: http://labs.adobe.com/technologies/stratus/


I think it's refreshing because it will soon support intranet p2p, both live and on-demand.

On-demand p2p is easier than live p2p.


"...enable publishers to dramatically reduce bandwidth costs by outsourcing media distribution to users."

That is a very bad title. All it does is shift the bandwidth costs from the publishers to the users. BBC tried to do this and caused problems for their users. BBC was surprised by the problems they caused, everybody technical said "we told you so."


Stratus has been around for over a year on Adobe Labs. P2P video and audio was introduced in Flash Player 10. FP 10.1 introduces multi-cast P2P and the ability to send arbitrary data over the P2P connection.


How long till someone comes up with a Skype clone on flash?


There were a number of us that built a Skype clone right after the initial Stratus release. Video/audio/text chat works pretty well over P2P, but you still need a centralized server for the initial connection (Stratus) and a way to get around firewalls (STUN/TURN).


Awesome. What happened to it?


"Eliminate bandwidth costs" sounds like your ordinary, exaggerated, sensationalistic marketing trick - just air. I also wonder how this will fare in the US and the UK, where internet infrastructure on the upstream side of things is more or less non-existant.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: