High-End Trading Strategists See Cost Savings in Cloud Computing

Cloud computing is changing the world for sophisticated institutional investors – even those who have never used Internet-based servers to execute a trade or stress-tested a high-frequency trading strategy.

Novell Inc.'s Server Room

Justin De Vries, a Novell employee, works on a server at the headquarters of Novell Inc. in Provo, Utah, U.S., on Wednesday, Oct. 6, 2010. Novell is in talks with at least two buyers, including VMware Inc., to sell the software company in parts, the Wall Street Journal reported last month. Photographer: George Frey/Bloomberg ***Local Caption*** Justin De Vries

George Frey/Bloomberg

Cloud computing is changing the world for sophisticated institutional investors – even those who have never used Internet-based servers to execute a trade or stress-tested a high-frequency trading strategy. The result, according to experts, will be a lower cost structure for performing high-end trades and more widespread access to the computing power needed to develop and execute complex algorithmic trading strategies.

Cloud computing – on-demand self-service Internet infrastructure where you pay-as-you-go and use only what you need – is growing fast. Revenues in 2009 topped $56 billion for a 20 percent-plus increase from the previous year, according to technology research firm Gartner Inc., which projects the market hitting $150.1 billion in 2013.

What makes computing in “the cloud” so attractive to institutional investors is that it enables an end user to “rent” computing time from organizations with huge server capacity, such as Google, IBM, Salesforce.com, Savvis, Microsoft’s Windows Azure, Amazon Elastic Compute Cloud (Amazon EC2), and Rackspace Cloud.

Firms now have a choice. They can build and maintain their own data centers, which can cost $1 million even if they cover only two or three markets, estimates Ken Yeadon, managing partner of Thematic Capital Partners LLP, a London-based venture capital firm that specializes in trading infrastructure. Or they can use cloud providers’ servers to test new trading strategies, back test and run time series analyses, and even execute trades.

“High-frequency trading strategies start with a lot of data analysis and often have a long R&D cycle,” notes Yeadon. “This can involve very expensive computational processes. You might need 1,000 CPUs working together in conjunction as a supercomputer for a short period of time. With cloud computing, instead of building the data center infrastructure yourself, you can test the strategy as long as you need to, on demand, and if it doesn’t stack up, you just shut it down and stop paying for it. These types of strategy would simply not be commercially viable otherwise for anyone except the very largest market participants.”

Providers like Amazon EC2 tend to charge the user only for the time and capacity used, rather than passing on a percentage of their total costs – including maintenance, troubleshooting, security and other capabilities. For instance, the hourly charge to use one Rackspace server with a 620GB disk powering 8,192MB of RAM costs $0.96 per hour, or $700.80 per month – along with two basic charges of $100 a month and $0.12 charge per server hour.


The capacity and flexibility of these systems, gained from experience hosting social networking and other large, complex “retail” systems, is far greater than that of proprietary systems at even the largest banks. The result, says Yeadon, is that renting space and time on the cloud is many times less expensive than designing and building one’s own hardware infrastructure.

Trading technology and analytics on the cloud is still in the early stages, says Lloyd Altman, senior executive in the capital market practice at consultant Accenture. Early adopters include newer and smaller players – “the proprietary trading firm with four or five people” – as managed service providers that cater to them, including Fixednetix – in which Thematic holds a stake – and Thomson Reuters Elektron.

The same goes for trading-related software-as-a-service, or SaaS, applications that users can license and run on demand. Microsoft is getting into the act as well with the introduction last summer of PowerPivot, a new version of Excel that can process at least 100 million rows of data simultaneously and runs on the Windows Azure cloud platform. “Cloud computing is permeating the whole supply chain,” says Yeadon.

Larger institutions, which often have a considerable investment in their existing data infrastructure, are less likely to move parts of their trading operations to the cloud – at least in the near term. But some banks are already adopting cloud platforms for other aspects of their work, suggesting that they haven’t closed their doors to the idea.

“We will never buy another data center,” Michael Harte, CIO of Commonwealth Bank of Australia, said in a speech to the Committee for Economic Development in Australia last April. “We will never buy another rack or server or storage device or network device again. I will never let any organization that I work for get locked into proprietary hardware or software again.”

One solution some banks are looking into is internal or “private” cloud networks, which virtualize their own computing services. Much of the technology that goes into creating and operating external clouds is now available to large organizations through vendors such as Eucalyptus Systems.

Analytics, research, and testing of trading strategies are the parts of the process that institutions find easiest to migrate to the cloud at present: “the stuff you need to do to get the car to the racetrack,” as Yeadon puts it. Actual trade execution? Not to the same extent, experts say.

“Program trades involving baskets of stocks, possibly yes,” says Altman. “But these need to take place in milliseconds. Ultra-high frequency, algorithmically decided trades based on real-time price moves, taking place in microseconds, are not going to move to the cloud.” Nevertheless, Yeadon notes, there’s savings for managers who can break up baskets of trades, using managed service providers that specialize in low-latency, ultra-high-speed automated trading for the transactions that need it, while saving money on the rest by using vendors that access the cloud.