Darrin Machay, Architect at YJT Solutions, a Chicago-based technology consulting firm, recently spoke with Sarah Rudolph about some of the biggest technology challenges facing the options industry.
SR: How long have you been at YJT?
Machay: I’ve been at YJT over a year now. I remember the day I started here, September 15, 2008, because the markets had started going crazy — my first week here. I witnessed firsthand the intense, immediate response level to our clients, and it was pretty impressive.
SR: What brought you to YJT?
Machay: I wanted to get into a firm where I’d be working on complex problems in mission-critical environments, and the fact that YJT only works in the trading and financial industry really appealed to me. A friend from college had been working at YJT and loved it and said come and talk to them. So I interviewed with them, and they sold me on the firm. It was a great decision.
SR: What are some of the technological challenges that apply specifically to options trading?
Machay: The options world is very different from the futures world, for example. One difference is the sheer volume of quote data. The typical volume on all the [U.S. options] exchanges is 1.7 million messages per second. That’s enormous. When you add the issue of latency, with the rise of algorithmic trading, there are a lot of technological challenges. Latency is an issue on any market. If you’re a split second late, someone else gets the order.
SR: Isn’t one reason options are more complicated the fact that there are so many different options classes?
Machay: Yes – risk management is more complex because one, options are derived from an underlying asset, and two, there are several different options classes. The value of the options is based on time, and the farther out the option is the bigger the time premium. And how close or far the option is to maturity determines how much of an effect the underlying stock has on the price of the option. Options classes each behave differently. All this must be taken into consideration when determining downside risk.
In options you might process over a million messages a second and to be competitive, you need to do this with low latency. So market data handling is a big challenge.
SR: How is it an issue and what are some of the solutions to data handling?
Machay: Firms may need to consolidate data from several exchanges and incorporate it into their trading strategies, often to take advantage of arbitrage opportunities. Because each exchange sends data in a different format, many firms use connectors to “normalize” the data. This process makes it easier to work with multiple feeds, but it adds latency to market data processing.
One solution to that is a chip called an FPGA (short for Field-Programmable Gate Array). A number of providers offer such a product, including Exegy and Celoxica. Exegy, for example, connects to major exchanges and normalizes data; it also maintains a virtual order book.
Low latency firms might use their own software or hardware from one of those companies to extract data from markets and put it into a unified format.
Another type of technology is multicast. Multicast is a computer networking term that is an alternative to unicast. It enables you to broadcast data to many systems without having to do the work of connecting to each one. The server sends one data stream and the network delivers it to each subscriber, greatly reducing the load on the sender. One drawback of multicast is that its message delivery is not always reliable. Some network protocols will automatically resend a message if it gets lost, but that’s not usually the case with multicast. Proprietary protocols can make it more reliable. 29 West, for example, has a messaging product that provides reliable multicast messaging, allowing you to recover from lost messages.
SR: Do all firms need these technologies?
Machay: That really depends on the firm’s strategy. Most hardware designed to accelerate market data processing is meant for firms with direct connections to several exchanges, known as Direct Market Access. Although this minimizes latency, it’s not required for every strategy. Some firms choose to use a less expensive consolidated feed instead. Consolidated feed providers, like Activ Financial or Wombat, maintain connections to the exchanges, and normalize and clean up the market data for you.
Multicast technology is very common for distribution of market data. Most exchanges provide their quotes only via multicast. Some consolidated feeds give you the option to use multicast or unicast connections; unicast requires less network configuration, but uses more bandwidth and puts more load on the server sending the data.
Another big tech issue in options is options analytics – the calculation of implied volatility and the Greeks – Theta, delta, gamma etc. The Greeks measure how sensitive an option is to changes in its underlying parameters, and are important for risk management. One method of calculating them is the Black-Scholes model. The secret sauce for many options trading firms is how they calculate implied volatility. Volatility is important because it can be traded like a price. If volatility goes up, your option’s value will probably go up. If you can predict volatility, you can make money with options. Many firms use variants of the Black-Scholes model; some use others. If you find a way to determine implied volatility better than others in the market, you will make money. Calculating implied volatility requires solving a fairly complex equation, and many firms do it across thousands of symbols on a tick by tick basis. Some firms use specialized hardware, like GPUs or cell processors, to accelerate these calculations.
Also, in networking technology, gigabit networking is typical because of the amount of data that most trading systems need to handle. One gigabit is probably the current standard, but 10 gig is emerging as the next standard. It’s good for trading firms not so much because the pipe is bigger, but for every second, more data is getting transferred in less time. That’s lower latency. Cisco has a Nexus 5000 series 10 gig switch that is designed for low latency. It’s much faster than a typical network switch.
Another technology is Infiniband. It’s used by some trading firms because it is a system interconnect that shares some of the features of Ethernet, but it is actually faster than even 10 gig Ethernet. It’s usually used for tightly coupled clusters or grid computing.
SR: Do you have any hobbies?
Machay: Well, technology. I have a mini data center at home. It has networking gear, servers.
SR: Is it all work-related?
Machay: Not completely. There are a lot of interesting topics related to computer science to think about. For example, Queuing Theory. Whenever you have a lot of data coming into a system that can’t process it right away, at some point you have to queue data. How do you optimally schedule planes landing on several intersecting runways, for example? What’s the best runway design? There are many interesting ways to solve that problem.
As for non-tech related hobbies, I have been known to venture outside on occasion. I enjoy snowboarding and am working toward a skydiving license.