The Future of Trading
10 Top CIOs on Wall Street | Why Mutation Will Matter | Buyside Traders: Not Taking Orders Any More | Calculating Cost First, Trading Second | Trading Desk, in One Box | Changing the Minds of Star Programmers | 2015: The Always On Desk
Changing the Minds of Star Programmers
November 4, 2010
Early this year, Hirander Misra left his position as chief operating officer of Chi-X Europe to become chief executive of a supplier of systems that support algorithmic trading strategies. He thinks the industry is only at halftime - if that - in the sophistication of the mathematical formulas that are used to execute increasingly complex and "multilegged'' trading strategies.
And, says the CEO of Algo Technologies, based in London, going forward it will take more than just knowledge of programming to succeed in developing high-speed automated trading systems. In fact, the entire mindset of a programmer will change in the second half of the advance of algorithmic trading, over the next five years.
"The typical idea of a programmer that you can hire is changing,'' said Misra. "Because now actually your ideal superstar programmer is not necessarily just the guy who knows good C++ , because the software level is coming closer to the hardware."
What a superstar in trading technology will need is "an ability to bring the different layers closer together,'' he believes. "That can only be done by guys that have a good knowledge of the entire technology stack.
"It's effectively a guy who understands the network, the operating system, the different components of hardware that can make up a machine and how that can be configured and the different latency implications that that has,'' Misra says.
EVERYTHING ON A CHIP
With the constant collapsing of cost and size of circuitry, Misra says that one of the most frequent questions he confronts is whether trading firms couldn't put all their algorithms onto a single reprogrammable chip - and have the exchanges they deal with exist on the same chip.
Interestingly enough, this is a reality already being confronted by NYSE Euronext, the operators of the New York Stock Exchange, NYSE Arca and NYSE Amex. Its NYSE Technologies unit has created middleware that allows technology suppliers such as IBM and Intel to put all the functions not of an exchange but of a trading desk onto a single server ("Network in a Box," page 16). Similarly, a St. Louis technology company, Exegy, has put a complete ticker plant into a single computing appliance.
But Misra says the idea of putting all the functions of trading firms' algorithms as well as the matching engines of exchanges onto the same programmable chip is years off.
There are probably only about 500 software programmers in the world, he contends, that are experienced and skilled enough at reprogramming chips. "There's got to be a wider proliferation of training and skill sets" for putting trading and matching onto the same chip.
The difficulty shows one reason why rewriting software has, until now, been preferred over rewriting hardware. Speed. If you're tweaking algorithmic trading strategies overnight, you want implementation the next day.
"In a high-change environment,'' said Misra, "it takes longer to reprogram a piece of hardware or an individual chip than it does to reprogram and test software where you can write and test on the fly.''
Programming itself is getting more complex. The advance of low-cost commodity chips with multiple processors in them means that tasks are being spread across more and more processors with results coordinated across hundreds or thousands of threads of calculations.