P2P Internet Traffic : The Elephant

Since several years ago, peer-to-peer issues have been a central topic on IP networking. This has been pictured in the program of some top networking conferences e.g. INFOCOM, SIGCOMM, ITC, GLOBECOM etc. What actually makes P2P issues so important?

One possible answer can be seen in the graph at the right. According to Telegeograhpy, as of 2007 54% of the total Internet traffic is originated from P2P applications which are in the forms of video (61%), audio (11%) and other formats (27%).

Some other facts (taken from Telegeography):

• P2P traffic accounts for nearly 60% of Internet traffic (Sandvine)
• The share of P2P traffic varies from 10% to 70% depending on a carrier's customer base and geograpic coverage (TG's interviews with carriers)
• When TPB torrent network was taken down in Sweden, TeliaSonera's Internet traffic declined by 60%!
• P2P accounted for 50% of upstream capacity on Japanese networks in 04/2006 (Japanese Ministry of Internal Affairs and Communications)

Applications that may or may not impact long-haul networks:

• YouTube serves 70 mn low-res videos daily!
• Online Gaming: World of Warcraft reported 6.5 mn subscribers in early 2006
• Video Calling: Microsoft claimed that users of Microsoft Messenger engaged in 1.1 bn minutes of video calling during 01/2006 (how about Yahoo?)
• Sporting events which routinely streamed online (World Cup, major league baseball etc.)
• Grid Computing : Teragrid (10 Gbps connections)
• In 06/2006 more than 106.5 mn people (3 of 5 US Internet users) streamed or downloaded video (in total, nearly 7.2 bn videos were streamed or downloaded - comScore 09/2006)

Intra-Asia IP Traffic

This post is closely related to the previous one, except that here we discuss about Internet traffic not about bandwidth capacity. I found an interesting statement given by a REACH official:

"In 1998, 100% of IP traffic between countries in the APAC was transmitted via the US. In 2004 it was around 30% (estimated)."

I do not know what criterion is used by REACH to come to those figures. But pesonally, I feel it is highly possible to be true. In around 1995, I remembered that when I called "traceroute" from my university to any local ISP, it gave me several US-hops! At that time I was dreaming a local version of NSFNET in this country (yes sure, later I knew that this is the timeframe where NSFNET was decommisioned!).

Many people at that time only concerned on establishing connections to the center of the Net i.e. the US and did not care about anything else. But slowly the "message" is spread and more and more people understood the idea behind this new thing. ISPs were making both bilateral and public peering(*). The govt has updated the regulation for encouraging local connections. Is it enough to suppress the inefficiency? Maybe! But I believe that right now intra-Asia IP traffic that travel first to the US is still in the figure of 10% or even more.


(*) for local readers, please note that peering != "piring" ! :-D