It was a busy weekend for Brenda Hoffman and her team about a month ago. The CIO of the Toronto Stock Exchange was overseeing the deployment of HP Integrity NonStop servers that use the Intel Itanium 2 processor.
The hardware and software, which cost about $20 million, is designed to improve the throughput of its core trading engine, which handles transactions involving some of the largest equities in Canada.
And, now that the new system is live, Hoffman has to start getting ready to replace the entire thing.
The TSX, like stock exchanges around the world, is getting used to the increased automation of trading that was once handled by human beings.
A growing number of pension funds, mutual funds and other institutional investors are using advanced mathematical models for making transaction decisions in the financial markets. The strict rules built into the model attempt to determine the optimal time for an order to be placed that will cause the least amount of impact on a stock’s price. Algorithmic trading, as it’s known, also means that IT systems can respond to electronic information about a stock — including information from newswires like Dow Jones — before human beings even have a chance to read it.
As a result, the legacy systems of organizations like the TSX are being put under much greater pressure to handle automatically generated trades.
In response, the TSX is getting ready to launch a new core trading engine later this year, called Quantum, which will be designed with algorithmic trading in mind. Quantum was originally code-named Leapfrog, and Hoffman, who was promoted to CIO earlier this year, admits that it represents a huge jump for the company.
“You need all aspects of the technology team focused on the new technology. It’s not just about the application changing,” she says. “From infrastructure and operations to third party applications we’ve had to refocus our efforts internally. It is a big undertaking, no question.”
Despite the compute power it brings to the table, the Itanium 2-based NonStop server installation, which was part of an initiative called TSX Express, is in some ways only a stopgap measure. The TSX is referring to the new core trading engine’s deployment internally as the Quantum Revolution. In some ways, however, the TSX has already gone through a revolution. In the 1990s, for example, the exchange suffered a number of high-profile system crashes that called into question its viability as a global player.
Then, in April 2003, the TSX reported two failure-free years following a system overhaul. It has also formed a division that seeks to assist other problem-ridden exchanges.
The Quantum system was developed internally by the TSX’s own IT staff, which numbers about 200 people. It will run on HP blade servers using x86 processors using Red Hat Linux. The company is sticking with HP, Hoffman says, because of its experience with high-availability processing, “which is a difficult thing to replace when you move off of NonStop.” Red Hat, meanwhile, was the Linux distribution the TSX’s technology team was most comfortable with, and is already deployed in other areas such as its feed distribution and tickers.
“They’ve proved it can scale to a product set environment,” she says.
David Easthope, an analyst with Boston-based financial research and technology firm Celent, says exchanges such as the New York Stock Exchange (NYSE) and NASDAQ are not necessarily creating new platforms but integrating electronic communication network (ECN) products into their existing trading engines.
“To some extent they do have to change and adapt,” he says, “but it may not be a huge change. The trades in an algorithmic trading scenario may be coming rapid fire, but as long as they can keep up with capacity and speed, then that’s really all the exchanges need to worry about.”
Easthope says it is more important for stock exchanges to invest in smart order routers that can be used for multi-asset trades including derivatives. The TSX has already announced plans to set up a smart order router.
Hoffman says Quantum has already gone through benchmark testing that show it can handle 320 million messages an hour or roughly two billion messages per trading day, with a throughput speed of 100,000 messages per second. Order matching will happen in millionths of a second, and the system can fail over in seconds with no lost messages.
“We’re working with both the business and tech folks on a migration strategy that will reduce the risk of implementation,” Hoffman says, adding the migration is expected to continue over the course of next year.
“We’ll migrate the symbols from the existing engine to the Quantum engine in a phased way. It’s not going to be big bang.” She also says ECNs have limited function compared to what Quantum will achieve. “They don’t have a lot of the special terms, the odd lots, the specialized orders we have in Canada.”
According to Sang Lee, managing partner at research firm the Aite Group in Boston, stock exchanges are starting to take a closer look at the exponential growth of messaging they are starting to see. Some worry that algorithms may break down, or cause inaccuracies.
“In the U.S. market right now, there are a lot of conversations around capacity issues, driven by algorithmic trading,” he says. “Sometimes to figure out what’s happening, you’ll have people sending out orders to see where the market is at. When they see what’s going on, they cancel their orders. That leads to massive capacity issues.”
Right now, the TSX conducts real-time monitoring of applications based on five-minute or even five-second snapshots. That data turns into forecasts so that system adjustments can be made. The TSX has created its own tool that will be migrated to Quantum to handle those metrics and others, including CPU utilization. No matter how algorithmic trading evolves, Hoffman says Quantum, and the TSX, is prepared for the worst.
“The order-to-trade ratio will continue to go up, which impacts how much hardware and software you need to do the processing,” she says.
In other words, Hoffman may have a number of other busy weekends ahead of her.