Awesome Conferences

See us live(rss)   

Can you make the speed of light faster?

One technical issue that often plagues me is that you can't make the speed of light any faster. Network latency from NYC to Sydney is going to suck no matter what. Helping users understand this is difficult. Often it is equally difficult to make software developers understand this too. Many times people have asked me, sometimes seriously, if we could just make the speed of light faster.

There is one obvious way to improve the latency between NYC and Sydney: Tunnel through the earth. A direct route would be much faster.

However it looks like scientists are close to a more realistic alternative: use air instead of glass!

"The speed of light, about 300,000 km/s, is the speed light travels in a vacuum. In a medium such as glass, it goes about 30 percent slower, a mere 200,000 km/s."

Here's the full article: http://ars.to/15SOcj9

If you aren't sure of the difference between latency and bandwidth, ACM Queue magazine's George V. Neville-Neil wrote a good description.. Here's a video explanation from Digital Society. I'd love to work with an animator to make an educational video demonstrating the difference between latency and bandwidth.

Posted by Tom Limoncelli

No TrackBacks

TrackBack URL: http://everythingsysadmin.com/cgi-bin/mt-tb.cgi/1564

4 Comments | Leave a comment

Neutrinos would be even better. Trouble is building a receiver that isn't a cubic kilometre of ice.

Skip fiber-optics and "tunneling" all together, and switch to a networking protocol based on exchanging neutrinos... They'll happily go straight through the earth at light-speed.

And there are plausible alternatives to reducing the latency of the physical links... You could have a lot more smarts in the routers at both ends of the links, understanding TCP/IP, and proxying numerous protocol details, rather than mindlessly transferring unnecessary packets.

For example, if the router at the server ends store entires segments of a TCP exchange (part of a file download) in their buffer, the nearby router could send back a forged ACK (with minimal latency) to continue the processes. While that's happening, it's transferring the entire exchange to the router nearest the client PC, which then masquerades as the server, sending these packets, handling ACKs/NACKs, retransmits, and other details. The routers (acting as proxies) could also get more aggressive, like toying with different MTUs on the source and destination, or cranking up the TCP sliding window. The two routers might even benefit from transparently compressing some of the data at the source, and uncompressing it at the destination.

I'm fairly certain that the main cause of latency in these sorts of cases is not the speed of light. It's due to the large number of devices that these frames and packets have to traverse on these extremely long paths. If you laid dark fiber between NYC and Sydney and *only* had repeaters and amplifiers I doubt you'd experience enough latency to cause one to consider changing the laws of physics. But it's very expensive to do this so most traffic ends up traversing numerous networks and devices--each adding a little more latency. So I'm rather doubtful that speeding up the light a little is going to make much of a difference on its own.

You CAN do things like reducing the number of HTTP requests tied to the rendering of a web page. That's the sort of thing that takes latency and multiplies it such that a page may load locally in

Leave a comment