The potential for huge wireless network throughput improvements based on work done at MIT and other universities is being met with a mix of awe and harsh skepticism.
Face it, when someone says they can boost throughput on a given network from 1Mbit/s to 16Mbit/s you'll take notice. When they tell me and others in the Northeast that people can really stream movies on the Amtrak Acela train between New York and Boston -- where simply getting email is tricky -- I really take notice. The only question is, as the commercials used to say, whether it's real or it's Memorex.
Basically, the gains come from utilizing coded TCP to attack the problem of dropped packets due to network delays or congestion. It is described in a paper written by researchers from MIT and University of Porto in Portugal, and drawing on work done at those schools, as well as Harvard University, Caltech, and Technical University of Munich.
The work focuses on the little lost packets of wireless networks, which may account for 2 percent, 3 percent, or maybe 5 percent of traffic (in the case of the high-speed train). The throughput gains come not from preventing the packets from being dropped but from eliminating much of the back and forth traffic involved in finding those packets and confirming their eventual delivery.
An article in MIT Technology Review says, "Instead of sending packets, it sends algebraic equations that describe series of packets. So if a packet goes missing, instead of asking the network to resend it, the receiving device can solve for the missing one itself."
The intellectual property is being licensed through an MIT/Caltech startup called Code-On Technologies. According to the researchers, the improvements can be gained by developing applications that can sit on existing network hardware, even smartphones, so no routers or other hardware would have to be replaced.
Students testing the technology on the Acela were able to stream YouTube videos while other passengers struggled just to get online. Muriel Medard, a professor at MIT's Research Laboratory of Electronics and lead author on the paper, said that in tests at MIT throughput on a network was boosted from 1Mbit/s to 16Mbit/s.
Dipankar "Ray" Raychaudhuri, director of the Winlab at Rutgers University, told MIT Technology Review that the work by Medard and others "is an important breakthrough that promises to significantly improve bandwidth and quality-of-experience for cellular data users experiencing poor signal coverage." He noted that the technology could be widely available in two or three years.
However, critics question the throughput gains. In a back-and-forth in comments on the MIT Technology Review piece, readers challenged the idea that you could get a 16-times improvement by addressing the loss of 2 percent or 3 percent of packets. Others said that still bigger gains could be achieved by redesigning today's networks, and that the coded TCP solution might improve throughput but might not be enough to improve end-user experiences that suffer as much from slow servers as lost packets.
The article and the technical paper's abstract are worth a read, but be sure to check out the comments too.