Folks some some weeks ago I posted a topic related to the French anti-filesharing groups, who conducted some filtering equipment tests, it seems more details have emerged. My thanks for finding this article go to "IceCube" Over at Slycks, cheers IceCube
http://www.zeropaid.com/news/9409/P2P+ISP+Filtering+Test+Published%2C+Labels+Deny+Ensuing+CriticismThe primary source of the report.
http://www.internetevolution.com/document.asp?doc_id=148803Out of 28 software vendors, 5 actually agreed to start the test and only 2 allowed the results to be published. Internet Evolution describes the results as "not perfect"
Identifying traffic proved difficult at the best of times. Bear in mind that these were likely the two top performing applications. According to the study, the two vendors which actually agreed to have the results published had the following detection rates
P2P Protocol Detection Efficiency
Ellacoya E30 Ipoque PRX-5G
BitTorrent 82% 97%
eDonkey 97% 88%
Gnutella 76% 96%
FastTrack 1% 97%
MP2P 86% 96%
iMesh 0% 47%
FileTopia 33% 23%
WinMX 7% 0%
SoulSeek 1% 5%
DirectConnect 7% 78%
The following table shows the percentage of how efficiently P2P traffic was regulated:
P2P Protocol Regulation Efficiency
Ellacoya E30 Ipoque PRX-5G
25% Regulation
BitTorrent 88% 88%
eDonkey 36% 63%
Gnutella 83% 93%
FastTrack 27% 91%
MP2P 93% 92%
iMesh 0% 43%
FileTopia 32% 16%
WinMX 19% 0%
SoulSeek 0% 0%
DirectConnect 12% 63%
75% Regulation
BitTorrent 90% 100%
eDonkey 40% 67%
Gnutella 57% 63%
FastTrack 97% 78%
MP2P 92% 93%
iMesh 0% 97%
FileTopia 85% 4%
WinMX 0% 0%
SoulSeek 0% 2%
DirectConnect 24% 58%
But wait, what about encrypted traffic? This was also tested and the results were obviously less than stellar on the vendors part. eDonkey traffic completely fell off the radar for both vendors while only Ipoque PRX-5G could not detect even half of the RC4 encrypted BitTorrent traffic.
Then there's dumping the traffic onto MPLS L3 VPN. Interestingly enough, only Arbor/Ellacoya even had the capability in the first place, but the performance was above 90% for the 3 protocols tested. There's plenty of other interesting findings in the study overall and definitely worth the read. For some, it may only serve to confirm that ISP level p2p filtering is not ready for "prime time" given that the existing encryption has a tendency to circumvent the software already. This goes over top of the well known fact that once moves are made to filter out P2P traffic, many developers of the technology are more than willing to develop new encryption methods to circumvent any new measure that comes their way.
Whilst labelling us as a "nearly dead network" in this report it seems the superior WinMX design features have been very long lasting, for a protocol designed in 2001 to still be up and running without change and outgunning newer networks is nothing short of amazing.
This is surely a feather in WinMX Technologies Cap