BitTorrent CEO Eric Klinker made some dubious claims to IDG News Service’s Stephen Lawson this week.  Aside from telling Lawson that the FCC had no authority to enforce its Net Neutrality rules which I’ve thoroughly debunked, Lawson reiterated BitTorrent’s persistent claim that their new Micro Transport Protocol (AKA uTP) was friendly to the network and other applications.  But the field tests that I have run and subsequent conversations with BitTorrent VPs confirm that these “friendly” features are at best token gestures with few measurable benefits.

Stephen Lawson attributed the following claims to Eric Lawson (source):

“This is the principle behind Micro Transport Protocol, Klinker said. The system instructs BitTorrent to take up unused capacity on a network so, for example, it will hold back during a busy work day to let more critical applications maintain their performance, he said.”

It’s funny that I’ve never observed this behavior in the 4 years of using and testing the BitTorrent protocol.  In fact when I simultaneously use my broadband connection to download with BitTorrent and a web browser, BitTorrent manages to hog 91% of the bandwidth over HTTP (see figure 1 below).  This isn’t “unused capacity” that BitTorrent is taking; it’s bandwidth taken from the web browser by exploiting a weakness of TCP congestion control.

Figure 1 – uTP still hogs over 90% of the network over HTTP

Source: Analysis of BitTorrent uTP congestion avoidance

Lawson further claimed:

“When it (BitTorrent uTP) senses delays, it slows down, helping other applications maintain low latency.”

But again, this claim is outright false when we look at test data from the field shown in figure 2 below.  BitTorrent causes massive amounts of jitter whether other applications are trying to use the network or not.  The amount of jitter (which is a spike in latency measured in milliseconds) peaked well above 1000 milliseconds which is an eternity in the context of Voice of IP (VoIP) or online gaming applications.  So even though gaming and VoIP want very little bandwidth (less than 0.1 Mbps), BitTorrent will cause frequent disruptions to them.  These spikes in latency mean that voice communication and game play are suspended for up to a second or more which is a substantial disruption that can cut off a few syllables or words in a voice conversation.

Figure 2: uTP upload & download traffic causes massive jitter

Source: Analysis of BitTorrent uTP congestion avoidance
Note: Horizontal axis represent samples which were taken once per second.  Vertical axes represent the increase in ping time (round trip latency in milliseconds) above the baseline latency of 11 milliseconds.

The peak jitter threshold for most VoIP applications (e.g., embedded analog telephony adapters or hardware IP phones) is as low as 30 milliseconds.  Every time jitter goes above that threshold, one or more VoIP packets are dropped and each packet drop results in 1/50 of a second disruption in the phone call.  Several packet drops in succession might prompt a “can you repeat that” from a VoIP user.

When faced with these questions, BitTorrent Product Marketing VP Simon Morris pointed out that they have capped the jitter induced by BitTorrent upstream traffic (uploads) to 100 milliseconds.  Even if this is an improvement over the older BitTorrent protocol (which I’m doubtful since my tests showed no improvement in uTP over the older BitTorrent protocol), 100 milliseconds is well beyond the usable breaking point for VoIP and online gaming.  Furthermore, the jitter induced by BitTorrent downstream traffic (downloads) is entirely unconstrained and it can frequently go above 200 milliseconds and occasionally over 1000 milliseconds.  This harmful and selfish behavior of BitTorrent was a conscious design choice because they would rather have better performance for BitTorrent than be less harmful to other applications on the network.

Despite the fact that BitTorrent is well aware of these harmful effects and their ineffectual measures to improve the “friendliness” of BitTorrent, they continue to spread the word that their protocol is friendly to the network and other applications.  It’s about time that BitTorrent answers for this and the media needs to question these outright deceptive claims.