Netli, Linux Take Web to Warp Speeds - page 4
The Joys of Sublight
As mentioned earlier, most Internet transmissions use the TCP/IP protocol to get around. It's efficient, and it basically works. But what Netli does is take the TCP out of the equation and use its own Netli transport protocol (NTP) to deliver information across the IP-based Internet.
They do this with their service euphamistically called NetLightning.
Here's what happens, using the Atlanta-Tokyo route as an example. A request from Tokyo heads for the Atlanta server, which is now using the NetLightning service. Instead of reaching the Atlanta server directly, Netli's DNS redirects the end-user's transmission to what Netli calls a Virtural Data Center (VDC) located close to the end-user's company or ISP.
The initial request is now converted to NTP/IP and now flows through the same Internet pipes that everyone else uses. But, because its Netli's protocol over IP that is being used, packets are pretty much transparent to the rest of the Internet trafic, and, because of the nature of the NTP, very few packets are lost.
The request tranmission is not sent directly to the Atlanta Web servers but rather to an Application Access Point (AAP) close to the Atlanta servers. There, the NTP/IP packets are converted back to TCP/IP and then sent to the Atlanta Web servers. Once the pages are served, the page's packets do the whole thing all over again: off to the AAP for conversion to NTP/IP, then zip across the planet to the VDC nearest Tokyo, convert to TCP/IP once more, then sent to the end-user's ISP and then finally to the end-user's browser.
One would think all this passing back and forth would take up extra time, and it does. But the time costs are simply negligible. Now, a trip that took 6.55 seconds now takes .25 seconds in the first mile, .1 second in the last mile, and (allowing for just two round trips) only .4 seconds in the middle mile, for a grand total of .75 seconds.