Slashdot Mirror
Drop in P2P Traffic Attributed To Traffic Shaping
An anonymous reader writes "A new report based on data from 100 US and European ISPs claims P2P traffic has dropped to around 20% of all Internet traffic. This is down from the 40% two years ago (also reported by the same company which sells subscriber traffic management equipment to ISPs). The report goes on to say the drop is likely due to continued, widespread ISP P2P shaping: 'In fact, the P2P daily trend is pretty much completely inverted from daily traffic. In other words, P2P reaches its low at 4pm when web and overall Internet traffic approaches its peak ... trend is highly suggestive of either persistent congestion or, more likely, evidence of widespread provider manipulation of P2P traffic rates.'"
There may be a "market saturation" effect. I know people who were downloading gigabytes a month (maybe a week) of songs and videos, but in the past year or two they have tapered off. They've gotten most of the stuff they've wanted, and now are just listening to and watching it.
there has to be more to this. obviously the ISP's are very aware of P2P networks. They market this in commercials that say "download music at increased rates!" which are in context about purchasing mp3's but belie the fact that they provide infrastructure to P2P networks, and anti-IP scenes.
And im not saying that this is a bad thing...
... usenet usage has grown to 25% of all internet traffic. people move on (or in this case back) to safer technologies. the xIAA are targeting P2P users, so people move away from P2P.
what's traffic shaping got to do with it?
Much more likely people are rescheduling their P2P downloads to run outside of peak hours. I know my ISP (Virgin Media) throttles connection speeds during peak hours, so I schedule anything I want to download to run outside of those times.
There really hasn't been all that much worth downloading as of late. You can only download the classics so many times, the new content coming out just isn't all that good, be it games, movies or music. I'm sure we'll see a small up tick when the new Star Trek movie hits the underground though.
Even my CD collection is gathering dust, finally music streaming that just works.
"The report goes on to say the drop is likely due to continued, widespread ISP P2P shaping"
The data allows no such conclusion to be drawn. In fact, since all they've done is compared P2P as a percent of total traffic, it's probably more likely that the total traffic has increased.
"A week in the lab saves an hour in the library"
...is ofcource spam and porn.
Can we do traffic-shaping of spam?
If so I suggest this shedule:
12pm-8am: 100% drop
8am-4pm: 100% drop
4pm-12pm: 100% drop
So torrents used to compose 40% of traffic. Now it's 20%. What's changed in the last year?
* youporn.com and similar sites have popped up where they did not previously.
* hulu.com now exists.
That right there could easily cover 90% of people's media interests. Especially now that I'm not really into movies as much as I used to be (they suck more, and TV shows are, in some ways, getting better).
~/ssh slashdot.org ssh: connect to host slashdot.org port 22: too many beers
Fact is:
- absolute P2P traffic volume is not dropping, it's just very slowly increasing
- absolute amount of HTTP traffic nearly doubled since 2007, thanks to major increase in online video and direct download services
=> many people often use DD today instead of P2P for filesharing
=> P2P percentage sharply decreased, not the absolute volume though
The drop in traffic is easy to explain. Most distros nowadays have a NetInst option, where you can download a small CD to boot off, then download only the packages you need.
All that P2P traffic IS just "Linux ISOs", right?
I can't believe that in all those comments nobody mentioned the most likely reason for those numbers:
Encryption.
Most of the P2P traffic will be Bittorrent. All popular bittorrent clients allow to use encryption and random ports to prevent traffic shaping. Encrypted torrent traffic can - to my knowledge - not be detected by the ISP and is most likely counted as normal traffic in the mentioned numbers.
Maybe encryption is not very mainstream yet but the hardcore users will always enabled it (even when their own connection is not limited) because it will result in better speeds. So every encrypted gigabyte they used to download normally affects the numbers twice: it's one less gigabyte of counted P2P traffic and one more gigabyte of counted normal traffic.
On a sitenote: this is also the solution for those affected by traffic shaping: tell you torrent client to encrypt the traffic at all times and watch your speed go up.
I am involved with an Internet streaming site (AmericaFree.TV) and our traffic patterns follow normal Television "Prime Time" - i.e., traffic peaks at roughly 6:00 PM to Midnight in the evening. This happens in the US, Europe and Asia, and the local time zone pattern looks a lot like the "Consumer-Internet traffic" graph (# 2 in the original article). (Note that all of these graphs do not start at zero traffic, but some higher value, like 50%). In our case (long format video), there appears to be relatively little streaming from at work.
If you look at Craig Labovitz's previous's post, What Europeans do at Night, it appears that European Internet usage drops quickly after dinner time, but I would interpret these graphs a little differently - European traffic starts dropping at 10;00 PM, while US traffic starts dropping at Midnight. This roughly matches what we see, and also European TV viewing patterns (see pages 22 and 23 of this presenation). Of course, American TV prime time is pretty similar to Europe's. Putting all of this together, I don't think that streaming video is driving the differences seen by Labovitz.
An interesting corollary of all of this is that there is still substantial bandwidth available for P2P in the hours after midnight. Off-hours P2P use could triple and still not be more than the current day-time use.
You have a CD collection? Someone, help me pull him out of 2000!
Beautiful poem.
It's rare to find real art like this in a comment form.
best feed around
If I may be allowed to re-order things a bit:
A new report based on data from 100 US and European ISPs claims P2P traffic has dropped to around 20% of all Internet traffic.
The report goes on to say the drop is likely due to continued, widespread ISP P2P shaping
reported by the same company which sells subscriber traffic management equipment to ISPs
Sheesh, evil *and* a jerk. -- Jade
Encrypted torrent traffic can - to my knowledge - not be detected by the ISP
See http://www.shmoocon.org/2007/speakers.html for Rob King and Rohit Dhamankar on "Encrypted Protocol Identification via Statistical Analysis".
Here's a brief recap: by looking at {mean value, variance} of {packet size, interpacket delays} going {up, down} and packet entropy for a specific flow, you get a point in a nine-dimensional space. Encrypted protocols tend to cluster together.
So here's the ISP algorithm: Measure a flow, find its nearest cluster, guess that behind the encryption is traffic of the protocol belonging in that cluster. If bittorrent, kill.
Note that Rob & Dohit don't look at how many simultaneous connections you make. That also tends to give away P2P traffic.
So the ISP can see you're P2P'ing. They can't detect whether it's illegal, or who should sue you, but they can (probably) see it's bittorrent.