There are lots of different ways to deal with latency and bandwidth contraints, but some of them you might not appreciate.
In 1994, a typical web page was about 50 kilobytes in size and dial-up modems could transfer no more than three kilobytes a second. Under such conditions, Web sites were built to use few pictures, and to start loading in three or four seconds,with full rendering in 20 seconds.
These days, handshaking and other issues account for most of the experienced end user latency.
Round-trip latency for “handshaking” typically runs 700 milliseconds these days. In part, that is because of domain name server translations, even when servers are uncongested.
On the internet, the average latency for corporate websites in America is currently around 350 milliseconds, according to the Network Weather Report operated by the University of California, Los Angeles.
Google’s latency is 150 ms, Facebook’s 285 ms and YouTube’s 515 ms. Such latencies will have to come down considerably if the next generation of internet applications, such as telepresence, high-definition video streaming and remote surgery, are to fulfil their promise.
When shaving a millisecond off the time needed to execute automated trades can increase revenue by $100 million, there is plenty of incentive to build private optical networks with latencies approaching zero.