P2P and P4P: The Tangled Web We Weave

It was Monday morning, and Haiyong Xie was running late. His flight to Los Angeles had been delayed, and then he had to face LA’s beastly morning traffic. Xie, of Yale University, was on his way to the P2P Media Summit at the Hollywood Renaissance Hotel to take part in a panel discussion about the P4P Working Group (P4PWG). The group’s sponsor, the Digital Computing Industry Association (DCIA), was the host of the event.

The panel’s other participants, Doug Pasko of Verizon and Laird Popkin of Pando Networks, held their own until Xie arrived about halfway through the talk. When he did, his late entrance and the congested traffic that had caused it conveniently underscored what the P4PWG is about: easing traffic through better management. P4P focuses on the Internet, though, not freeways.

Not Just for Pirates

Peer-to-peer (P2P) networking decentralizes the exchange of data. Instead of an application retrieving data (for example, a large movie file) from a single, all-knowing server located deep in the bowels of some online entertainment distribution enterprise, a P2P app will connect to dozens/hundreds/thousands of computers across the Internet to simultaneously download bits of the file it doesn’t have and upload bits of the file it’s already acquired to other users.

Proponents of the technology say it prevents a single server’s pipeline from getting clogged with too much data at times of high demand.

Even rare data files that no longer exist on any centralized server might live on, constantly replicating on P2P networks.

Even though video-streaming sites like YouTube are taking up an increasingly large portion of Internet traffic, P2P traffic remains a substantial portion of all Web activity — 37 percent, according to an Ellacoya Networks study from last June.

Since Napster’s heyday in 1999, P2P file-sharing continues to be the pirate’s tool of choice for obtaining and distributing copyrighted material. However, legitimate organizations are warming to the concept of P2P as well. Above-the-table operations like Joost, Skype and Microsoft’s Threedegrees use the technology. The DCIA, which promotes P2P, counts Cisco, Verisign and AT&T among its members.

Later on Monday, at another panel discussion at the same DCIA conference, the argument was made that large entertainment companies that distribute video online will need to embrace P2P technology more firmly in the face of growing demand for high-bandwidth content like high-definition movies and TV shows. The decrease in prices for ever-wider bandwidth is not enough to address the problem, contended Michael King, CEO of Abacast.

“Broadband prices are dropping, but the drop is slowing down,” he commented. “Demand for high-quality content is speeding up. So the argument that nothing’s needed is wrong.”

P2P has shortcomings of its own. In choosing which fellow sharers to hook up with, a P2P application will sometimes make connections across networks, across ISPs and across the world. The result is a tangled web of connections that do the job of sharing data but not always in the most efficient way, at least not from a typical ISP’s point of view. These sorts of unregulated hook-ups are part of what led to Comcast’s practice of throttling back P2P traffic, which drew outcries from Net neutrality advocates.

This tangle is what P4P intends to address. The system introduces an element of logic to a P2P network that directs user traffic to first share with neighboring computers on the same network or in the same region before attempting to hook up with computers that are farther away and necessitate connecting with other networks. It’s distributed computing’s answer to the locavore diet. Its intention is to raise efficiency and lower costs for the ISP.

The panel said that tests run with Verizon yielded promising results. Whereas 98 percent of the traffic on a “naive P2P” network was external, implementing P4P cut external traffic to 50 percent. Under P4P, 29 percent was local, and 21 percent of traffic was internal, they said.

Questions Remain

With all the sunshine the DCIA is pouring on P4P, one must wonder what the downside is or what controversy might be involved. The panel members appeared to anticipate such questions — they even worked a few “myths” about P4P into their slide presentation.

The question of privacy was raised: How much will the management applications know about who a sharer is and the exact nature of the data transmitted and received? According to Pasko, it will be quite minimal. “A little bit of information goes a long way.”

However, a section of Pando’s site that explains the P4PWG states that the group’s goals include promoting ways to improve P2P “while protecting the intellectual property (IP) of participating entities.”

Bringing the term “intellectual property” into the discussion raises the question of whose IP they’re talking about. Does the term “participating entities” refer to companies like the ones in the P4PWG’s “Core Group” of partners — BitTorrent, LimeWire, etc. — whose patents it means to protect? Or does it mean content providers in the “Observer Group” — NBC Universal, MPAA, etc. — whose copyrights it means to protect?

Defending one’s copyright against pirate file-sharers is a legal right, of course, as is protecting one’s patents. But if in doing that the P4PWG intends to inspect the data being transmitted for copyright violations, that’s not exactly a “little bit of information.”

Traffic Management

Another issue facing P4P is the suspicion that the technology is merely a way for ISPs to avoid a head-on collision with Net neutrality legislation that would legally prevent them from managing traffic at all. It’s more than a suspicion, really; Comcast VP Kathryn A. Zachem all but said so in a filing with the Federal Communications Commission last month following a P4P test conducted by Pando.

The test, she said, “provides further proof that policymakers have been right to rely on marketplace forces, rather than government regulation, to govern the evolution of Internet services.”

The government regulations to which she’s alluding — if, hypothetically, they’re enacted — may well forbid the very kind of network management P4P represents. Apparently, Comcast would rather have a little network management control than further risk provoking a law that would guarantee it none at all.

How much control does P4P entail? At the DCIA panel, an audience member asked a question regarding peer preference. If User A finds he can connect with User B at a fairly fast rate, but she resides on a different, far-away network, will P4P force User A to connect instead with User C, who is on a closer but far slower connection?

Popkin’s answer was that using one would not prevent you from using the other; you’d just connect to the closer one first. “We’re not going to force you on a communication path that’s not going to provide the service,” added Pasko. Details were not provided regarding exactly how slow a nearby peer would have to be before a P4P system would no longer prefer that connection.

The panelists’ point was that farther but faster connections won’t be cut off; they’ll just be forced to the back of the line. Remember, though, that when Comcast throttled back BitTorrent transfers, it didn’t cut them off completely, either. They were just essentially told to get to the back of the line, again and again, until users gave up.

Click here to e-mail Paul Hartsock.