Zazzle Shop

Screen printing

Wednesday, November 5, 2008

Comcastic P4P trial shows 80% speed boost for P2P downloads

By Nate Anderson |

Comcast engineers have just released the first-ever real-world data on P4P technology—and it appears to be a massive success. While only a trial, the results do show that P4P's iTracker technology can increase P2P download speeds by 80 percent on ISP networks without materially increasing the network load.

P4P, which is being designed under the aegis of the Distributed Computing Industry Association, is meant to "localize" peer-to-peer transfers. P2P users generally grab data from all around the world, putting tremendous cost and bandwidth pressure on ISP peering and transit links with other networks. P4P uses an iTracker server to keep those transfers within an ISP's own network when possible, with the goal of boosting speeds for users and lowering peering-point loads for ISPs.

(Note that P4P has little to do with network neutrality, though companies sometimes suggest that it does. While it can decrease the financial hit ISPs take from sending traffic over transit links, P4P isn't designed to lower the total bandwidth on an ISP's network. In fact, by directing P2P users to "local" network sources, it might actually be expected to increase an ISP's local network traffic. More on that below.)

Comcast engineers have just filed the results of the first major P4P trial as an "Internet draft" with the IETF. The trial involved Pando, Yale, three (unnamed) ISPs, and Comcast, and it took place over the summer. It used a special, Pando-provided P2P client that is set up to check in with "iTracker" servers when searching for download locations in a BitTorrent swarm. The test used a 21MB video file (which was "licensed," in case you were worried), and measured the results of using the P2P client in order to see how the use of iTrackers affected uploads and downloads.

Data source: Comcast

Results were hugely positive. Compared to a random swarm, the use of any iTracker provided substantial speed boosts to Comcast network users, ranging from 57 to 85 percent above default behavior. For consumers, this would obviously be welcome news, but how does it affect Comcast?

Not too much, it turns out. "We did notice that download activity in our access network increased somewhat, from 56,030MB for Random to 59,765MB for P4P Generic Weight and 60,781MB for P4P Coarse Grained," wrote the Comcast engineers. That's a small increase, especially given that it reduced Comcast's "incoming Internet traffic by an average of 80 percent at peering points."

But uploads proved even more surprising. "It did not appear that P4P significantly increased upstream utilization in our access network," note the engineers. "In essence, uploading was already occurring no matter what and P4P in and of itself did not appear to materially increase uploading for this specific, licensed content."

Further good news came from a close examination of the various iTrackers. The fine-grained tracker provided detailed topology information about Comcast's network, and it took significant time to set up. "It was a detailed mapping of Comcast backbone-connected network Autonomous System Numbers (ASN) to IP Aggregates which were weighted based on priority and distance from each other," says the report. "Included in this design was a prioritization of all Peer and Internet transit connected ASNs to our backbone to ensure that P4P traffic would prefer settlement free and lower cost networks first, and then more expensive transit links."

Sounds like a lot of work. But Comcast found that it wasn't even necessary; simpler iTrackers provided even better results. The best results came from the coarse-grained iTracker, which featured 22 iTracker node identifiers and "resulted in a 1,461 line configuration file."

iTrackers can run on lightweight servers; for good results, every ISP should run one. While every network that deploys an iTracker will see good results, the system gets "dramatically" more impressive as more ISPs deploy it.

Comcast wants the iTracker mechanism made an IETF standard; more impressively, it recognizes that iTrackers should be "opt-in" mechanisms that each user can control. "The use of ISP-provided iTrackers should probably be an opt-in feature for P2P users, or at least a feature of which they are explicitly aware of and which has been enabled by default in a particular P2P client," say the engineers. "In this way, P2P users could choose to opt-in either explicitly or by their choice of P2P client in order to choose to use the iTracker to improve performance, which benefits both the user and the ISP at the same time."

Win/win scenarios haven't been common in the P2P wars, and it's important to note that this was only a preliminary test. But if P4P shows the same promise in larger trials and if P2P clients start to embed the technology into their software, faster downloads and lower-traffic peering points could be had simply by running some inexpensive servers.