A Little Fancy Maths Could Make Mobile Networks Five Times Faster

A Little Fancy Maths Could Make Mobile Networks Five Times Faster

LTE might be fast, but it sure ain't fast enough. Enter new research from MIT and Caltech, which suggests that a little fancy maths could boost mobile data transmission rates -- by as much as 400 per cent.

Network World reports that the collaboration has given rise to a new data transmission method which used something called Random Linear Network Coding. Unlike usual transmission techniques, RLNC -- as its best buddies call it -- encodes each packet of data being sent using information from the previously sent packages, and a few randomly generated coefficient thrown in for good luck, using some linear algebra.

That may sound uninspiring, but what's clever about it is that it can recover from errors without the sender or receiver ever retaining transmission information or having to request packets to be resent. How? Well, it simply works out what the missing packet contained from a later-sequenced packet -- that, by definition, includes earlier-sequenced packets and the coefficients used to encode the packet.

Big deal, huh. But remember that a big reason for slow mobile data is the way the networks deals with missing and corrupt data. With RLNC that's just not a problem, despite the fact that packets are necessarily a little larger than usual -- and it's borne out in testing. Pitting conventionally encoded Wi-Fi with RLNC-coded Wi-Fi, researchers looked to see how quickly a four-minute video could be downloaded when it had a 3 per cent error rate, which is not uncommon. RLNC encoding was five times faster than normal Wi-Fi. Five. Times. Faster.

Excited? You should be. RLNC has the added benefit of allowing the use of 4G LTE and Wi-Fi data streams in channel bonding applications -- in other words, it can double up data streams to use them both, for the same data transfer, at the same time. And because it's all maths, it can implemented entirely in software, with no need to upgrade hardware.

The only bad news is that it's still lab-based. Hopefully MIT and Caltech change that soon. [Network World]