Applied thinking

When Evrard van Hertsen sold Renaissance Software, his successful interest rate derivatives systems company, in 1995, he went looking for the next new thing in financial technology. This was in Silicon Valley in the early days of the internet era, with the dotcom boom about to take off. But what van Hertsen was looking for was not a virtual new economy concept, but a real solution to the problem of building trading and risk applications that could keep pace with finance innovations, that could operate at the speed of electronic markets and that could support all asset classes on a single platform. “It had been the goal of every software application builder, whether vendor or internal group, to do this,” says van Hertsen.

There were some technologies that could, in theory, do the job, such as object request brokers and other object-orientated approaches. But they were complicated and time-consuming, and it was difficult to add new functionality. Then as the web took off, so too did a raft of new technologies.

Chief of these was application server architecture – a way of breaking systems down into components and distributing them across multiple machines, where they could provide computing services to whichever applications or users needed them. Van Hertsen reckoned he could take the concept that was initially used to provide relatively simple functions on a server for thousands of users, as in internet applications, and turn it on its head to provide highly complex, computationally intensive financial functions for a relatively few users per server. This would solve problems of performance and the reuse of common functions across applications. Investment bank IT departments were also monitoring the technology, but it was all new stuff then and there was a risk in adopting it. “In Palo Alto [in the heart of Silicon Valley] we knew it would work, but on Wall Street they couldn’t be sure,” he says.

This gave van Hertsen and his new company, Application Networks, which he set up in 1998, a head start in building a cross-asset trading and risk management platform. Investment capital for high-tech ventures was easy to come by then, but Application Networks’ approach was somewhat against the grain of the times. While most dotcoms burnt through millions of dollars creating and marketing ever-more fantastical concepts, Application Networks was a heads-down engineering team that spent three years out of sight building its new architecture. It was a major undertaking that took $15 million and 40 experienced financial and software engineers based in Palo Alto and London. They used other emerging technologies, particularly the Java development tools, caching techniques whereby large databases can be held in memory for instant access, and low-cost server machines that make it cheaper to have multiple processors for performance and to build in the redundancy necessary for reliability.

Application Networks created a three-layered framework, which it calls JRisk (the ‘J’ is for Java), with a set of common services at the bottom, such as fault tolerance for reliability and distributed computation for performance, that would be required by all applications. Above that is a layer of reusable business logic components for representing instruments, handling market data and so on. And on top of that sits the trading and risk applications themselves. Initially, Application Networks left this top layer for its clients to build their own applications, and the company’s first few sales were for the first two layers only.

“We sold to investment banks that usually only did in-house development,” says van Hertsen. But then the dotcom bubble burst and in 2001 the economic climate got difficult. “We realised specific business applications would make it easier to sell our framework,” he says. So the company began developing trading programs, starting with a credit derivatives application.

“We said we would deliver the application in six months,” says van Hertsen. “That’s something we could never have done before.” Application Networks’ framework, with its set of common services and reusable business components, meant the company had to develop only the functionality that was new for credit derivatives, such as descriptions of credit instruments, some specific credit risk measures and certain elements required for processing credit default swaps, collateralised debt obligations and other credit instruments.

“Our biggest achievement is that we have been able to develop applications using our basic framework for instruments we never dreamed of,” says van Hertsen. This ability to rapidly develop new applications is a boon to IT departments. “As one chief technology officer put it: he wanted to be able to run ahead of the user requirement curve instead of struggling to keep up,” says van Hertsen.

Application Networks now offers applications for equity, interest rate and foreign exchange derivatives trading and profit and loss calculation, as well as credit derivatives. Meanwhile, clients, of which the company has seven, including Société Générale and JP Morgan Chase, have built applications for interest rate, equity and forex derivatives and treasury on the JRisk framework.

And once they have built an application for one asset class, it is quick and easy to build applications for other asset classes, claims van Hertsen. “We wanted as much reusability as possible so that when you build a new application you don’t have a lot to reinvent.”