Goldman also has developed an architecture that enables the automatic provisioning of servers, which, Squeri says, has enabled the firm to take advantage of consolidation opportunities in its data centers and reap significant capital and operating expense savings, as well as to manage the provisioning of servers more efficiently. The servers are virtualized with VMware; Goldman built the management layer that operates these virtual servers.
This year the firm also moved to thin desktops for all elements of its businesses. According to the firm, this has reduced power consumption and provided flexibility in the area of business continuity while enabling employees to work remotely. Squeri notes that some legacy systems needed to be tweaked to work across the thin desktops, but otherwise the process has gone smoothly, he says.
The IT focus for next year at Goldman will be on "leveraging components we've built in the past for algorithmic and automated trading across multi-asset strategies," Squeri says. "We're building automated trading engines for our in-house traders as well as our external client trading systems because we feel the biggest innovation we'll get over the next 12 months will be the ability to move toward more electronic trading across multiple asset classes."
This is a somewhat astonishing statement coming from an executive at the firm that dominates the realm of automated trading. Surely equities are already traded electronically at Goldman?
"They are, but right now for many of our external clients equities are still traded in somewhat of a manual fashion," Squeri admits. "If you've got a list of stocks you need traded, you wait for the market opportunity, submit an order down to the exchange or to a simple algo, wait for your trade fills to come back, then submit the next order. We're building automated multi-asset trading strategies that allow you to put much broader trade ideas into the algorithm and have the algo trade it for you through electronic pipes."
Like other high-frequency trading programs, the software takes in a variety of signals, such as market news and trade ticks, reacts to those signals, and adjusts or implements automated trades, explains Squeri, who adds that Goldman is building new trading algorithms for several areas of its business, including fixed-income products and cross-asset-class trades. [This project made major news in July, when Sergey Aleynikov, a former VP at Goldman Sachs, was arrested for allegedly running off with some algorithmic trading code.]
Also on Goldman's plate for 2010 are projects to provide enhanced analytics to clients through its front-end Goldman 360 platform. "A lot of it is about enabling our clients to operate and execute more efficiently, letting them benefit from some of the technical innovations that we've benefited from in-house over the past three years," Squeri comments.
The firm has developed technology that lets it expose interactive models and real-time analytics to external clients via a Web 2.0 infrastructure. When these projects are completed, clients will be able to see how some of their trading algorithms have performed and get a better sense of which ones to continue and which ones to drop, Squeri notes. "We have the ability to do that because we have a common platform across our internal and external systems with physical barriers separating the two," he says, explaining that while there's a shared platform, two different broker-dealers are involved -- one for internal and the other for external clients -- so the two environments are physically separate. "That's difficult to architect," Squeri adds.