Who wouldn’t want to watch streamed video over a Wi-Fi connection without the dread “wheel of waiting?” Or 20-times faster satellite data throughput? All without an overhaul of the underlying networks themselves?
At its most basic level RLNC relies on an algebraic equation to improve data speeds by reducing dropped packets. Those misplaced data bundles lead to congestion across a wireless network as devices try to recover the missing pieces. What the equation does instead is describe the packets, and somehow if a packet gets lost in transmission, the receiving device can solve for the missing piece without getting bogged down. (It sounds like magic to me, but I hope to learn more from the company’s presentation at Structure on June 18.)
As Partridge (pictured above) described it, the way things happen now is data gets chunked up and sent over the network and, with TCP, those chunks have to be received in order. Code On, however, uses linear algebra to combine packets and represent each packet as an equation. No compression is necessary. And the packets can be received in any order.
Network World published a pretty good explainer:
In over-simplified terms, each RLNC encoded packet sent is encoded using the immediately earlier sequenced packet and randomly generated coefficients, using a linear algebra function. The combined packet length is no longer than either of the two packets from which it is composed. When a packet is lost, the missing packet can be mathematically derived from a later-sequenced packet that includes earlier-sequenced packets and the coefficients used to encode the packet.
The upshot is that by applying a small bit of code “anywhere from the chip to the application layer to base stations” Code On can boost the performance of Wi-Fi or cell networks. “We can make even crappy wireless networks” 30 times faster, Partridge said.
In a demo Code On likes to show, it takes 5 minutes and 34 seconds (and 13 spinning wheels) to download a 3:54 video on a “good” Wi-Fi connection; i.e. one with three percent packet loss. With Code On applied, that same download on the same network takes one minute and five seconds (and no spinning wheels.)
In this example, the technology would be deployed in an encoder at the proxy server and a decoder in the form of a downloaded Android or Apple app on the mobile device.
There could be broad application of RLNC in cloud technology as well. Use of the technology could allow much more secure storage of files that can be cut up and stored across multiple clouds. “You could put parts of your file in Dropbox, in OneDrive(s msft), in Google(s goog) Drive and Box,” she noted. The beauty of RLNC, is you don’t need to store all coded or encrypted files in one location but instead put portions of those files in multiple locations and encrypt just the coefficients of the equation,” she noted.
The company’s underlying technology was invented at MIT and Caltech labs and its founding technologists are: Michelle Effros and Tracey Ho of Caltech; Muriel Médard and Dina Katabi of MIT; and Frank Fitzek of the University of Aalborg.
Code On’s goal is to push for broad adoption of its proprietary technology via licensing to third parties. The universities need to be paid for their work but the goal is to propagate the technology widely. Said Partridge: “From our perspective, we license out the core technology but unlike Qualcomm(s qcom), we don’t enforce how those companies use it. We want them to go out and innovate.”