P2P – Boon, Boondoggle, or Bandwidth Hog? – The Dark Side

Yesterday’s post explained how peer-to-peer (P2P) applications use the processing power, bandwidth, and storage capacity of participants in a service rather than centralized resources. This makes such applications generally less subject to catastrophic failure, much less subject to running out of resources (since each new user brings new capacity as well as new demand), and much cheaper FOR THE PROVIDER of the application in terms of hardware and bandwidth required.

It’s the FOR THE PROVIDER part that’s the rub. Let’s consider the case of BBC’s iPlayer service. For a seven days after most broadcasts, UK residents over 16 years old can download the show free and store it 30 days on their PCs for later viewing which can be offline. The current version doesn’t even download ads with the shows.

Sounds great, right? Just what TV should become on the Internet. Not so fast, according to British ISPs. They complain that this new service will overload their networks and that BBC has no right to do this (although the British regulator has given them permission). The ISPs say they may have to throttle the number of people who can get the shows or “protect” themselves in some other unspecified ways. Usually (by Fractals of Change, at least) the UK is held up as a model of a competitive market for broadband services where issues of net neutrality don’t raise their ugly heads. What’s going on here?

I suspect that the fact that iPlayer is a P2P service is at the root of the problem that ISPs have with it. If BBC downloaded shows to those who requested them directly from its own servers and if the service proves as popular as it might, BBC would have to buy huge (or huger) pipes of its own into the Internet and ISPs would get some revenue from that. However, in a P2P implementation, it is likely that only a few seed copies will be downloaded directly from BBC to the first people who ask for a particular show. Subsequent requestors will get their copies from the first requestors. BBC describes it this way in their terms of services (TOS):

“When you install the BBC iPlayer Library you will also install peer-to-peer file sharing software from Verisign Inc. This software has a file share feature that enables other BBC iPlayer users to download BBC Content through your personal computer (using part of your upload bandwidth), via a secure link, to their personal computers. … When you use BBC iPlayer Library you shall not have the option to 'switch off' the peer-to-peer functionality as this is a core component of the BBC iPlayer Library.”

BBC also warns: “… you are responsible for paying all expenses that you may incur in connection with your access to and use of BBC iPlayer including your internet service provider charges and any excess charges to that provider if you have a cap on downloads and/or uploads…”

OK, fair warning if you’re in the habit of reading TOS carefully. But this warning does NOT appear in any of the marketing information for the service which I saw.

So why should the ISPs be upset? If the users upload too much, they’ll be charged more and the ISPs’ll get paid more – not by BBC but by the users themselves. The problem is that not all ISPs have upload or download caps. Those that do usually don’t advertise them very prominently if at all. Moreover, most users may be well below their caps now – almost certainly are. So the users will use bandwidth which, from their point of view has been sitting idle, and either won’t pay the ISPs any more or will be outraged if they are presented with a bill or thrown off an “unlimited” service for overuse.

It is highly unlikely that you use more than a small fraction of the bandwidth on your connection to your ISP most of the time – usually your connection is idle. But ISPs count on the fact that most connections are idle most of the time; there isn’t nearly enough backbone capacity to handle all the traffic which would result if all the local connections were busy all the time.

“Oversubscription” isn’t fraud; it’s the correct way to design networks which are inexpensive enough for their users to use. The phone network wouldn’t work if all phones (or even more than a small fraction of phones) were offhook at the same time. The highway system wouldn’t work if every driveway were disgorging and engorging its maximum capacity 7x24. The backbone of most networks is designed assuming that most spurs will be idle most of the time.

So if BBC succeeds in low cost distribution of its content without buying new capacity of its own and users substantially increase their use of download and upload capacity without paying extra themselves, the ISPs face either providing degraded service or a sudden need to upgrade their networks (or higher charges for uses of other providers’ networks).

Are the ISPs right to try to make BBC pay directly or indirectly for use of enduser bandwidth? Next post a smart reader suggests that Internet access be metered to avoid even the perception of P2P arbitrage at the expense of ISPs.

Timely note: Ironically, as I write this, P2P network Skype is experiencing a rare outage. P2P services are generally more outage resistance than services which depend on a centralized bank of servers which can be subject to all sorts of catastrophes but they’re not invulnerable. Some networks which perform most of their functionality between peers may still rely on a central server to coordinate. Also new software is just as likely to act up in an unanticipated situation when it’s run on a distributed network as it is when it’s run at a central site – this appears to be Skype’s problem; in fact, it’s a bit harder to back out a bad upgrade on a P2P service than a centralized one.

Comments

[Steve]

Phone off hook or on hook. PC power on or PC power off. P2P request or No P2P request.
I think that usage or bandwidth requirements are not going to change. Just like a phone call, if there isn't a P2P request by the requested PC, then there won't be any demand for any bandwidth. All expectations for a certain number of Idle PC's will remain the same. There is a reason there are idle expectations built into bandwidth requirement formulas. Cause, Most people are busy doing something else. The ISP concerns might be realized someday in the coming next decade, but today they are unfounded. Just another reason to try to raise your cost per internet connection.

[Steve]

Phone off hook or on hook. PC power on or PC power off. P2P request or No P2P request.
I think that usage or bandwidth requirements are not going to change. Just like a phone call, if there isn't a P2P request by the requested PC, then there won't be any demand for any bandwidth. All expectations for a certain number of Idle PC's will remain the same. There is a reason there are idle expectations built into bandwidth requirement formulas. Cause, Most people are busy doing something else. The ISP concerns might be realized someday in the coming next decade, but today they are unfounded. Just another reason to try to raise your cost per internet connection.

[HandsOffPlease]

I'm interested to read your follow up posts. Either way, there is great potential for the market to work itself out. Consumers' collective demand will have control. As was the case with ESPN360 reported by Renesys.
http://www.renesys.com/blog/2007/08/espn360reverse_net_neutrality.shtml

[Aswath]

Usage based pricing (both for upload and download) model seems to be an equitable solution to the problem. Granted, users need to be retrained from the flat-rate pricing, but it eliminates ISPs' cries of wolf.