In the past few days, there’s been a lot of talk in the news about latency. Working at a company that makes network simulators (www.apposite-tech.com), I spend my day talking about latency, but still I was not prepared to wake up to the sight of Cramer howling about latency on CBNC this morning.
For those of you who have been too busy fighting fires on your network to follow the financial news, here’s a brief recap: The WSJ and CNBC this week have been covering a report by Eric Hunsader (http://www.nanex.net/aqck2/4436.html) that shows that high-speed traders were making huge trades on gold, stocks, and bonds within a millisecond of the Federal Reserve’s surprising announcement last week that they would not scale back their purchase of bonds.
High-speed traders regularly buy and sell based on news in a matter of milliseconds or less. No surprise there. However, the news was supposed to be released from Washington at exactly 2:00 PM EDT as measured by the official USNO atomic clock. For the information to reach Chicago from Washington takes about 6 ms. (For more gory details, see my article on latency: http://www.lovemytool.com/blog/2013/08/its-the-speed-of-light-dudeby-dc-palter.html)
The speed of light through fiber is approximately 5 ms per 1000 km, and the two cities are approximately 1200 km apart. If someone was transmitting the information through the air using lasers or radio waves, it would get there in about half the time, but it seems impractical. So for trading to start in Chicago at 1 ms past 2:00, the information must have been sent from Washington more than 5 ms before 2:00. And that is not supposed to be happen.
So either someone figured out how to break Einstein’s law, or the information was distributed early. (If someone did invent a way to transfer data instantaneously, it would be far more lucrative to trade secretly on financial news than make the discovery public and accept the Nobel Prize, but that still seems unlikely.)
If the big trades had started at even 1 ms before 2:00, it would have been obvious that they had the information early. So I suspect that whoever had the information thought they could start trading at 1 ms afterwards and nothing could be pinned on them. It was only because Eric Hunsader understood the issue of latency of data transfer over the WAN that anyone noticed that something was amiss. Unless an explanation is discovered quickly, this will probably become a topic of Congressional hearings and we’ll get to hear not only Cramer but all of Fox News and MSNBC screaming about latency.
If you want to see how 5 ms of latency will affect your applications, or simulate a link between Washington, Chicago, and New York, get one of our Linktropy or Netropy network simulators (http://www.apposite-tech.com/products/).