The Gateway to Algorithmic and Automated Trading

AT webinar: Tech execs see ways to boost performance while cutting costs

First Published 30th May 2014

Looking to improve performance and reduce costs without redesigning your entire system? A range of tech experts in an Automated Trader webinar discuss ways this can be achieved.

Gil Tene, CTO, Azul Systems

Gil Tene, CTO, Azul Systems

"Being fast but taking 10 years to get to the market is useless."

Developers looking to reduce latency in trading applications would love to redesign from the ground up, but what if that's not an option?

Mark Skalabrin, Redline Trading Solutions' CEO, said that one of the biggest challenges for developers is balancing the pressures of improving trading performance with the need to lower operational cost while making decisions over which components of legacy systems to keep or change.

Skalabrin was speaking in an Automated Trader webinar hosted by Intel, which also featured executives from Azul Systems and IBM.

"For a lot of reasons it is not always possible to redesign a legacy application. The next best thing is to maintain the current interface that application uses but replace the machinery that it uses to get to the market," he said.

Redline provides an OpenMAMA bridge that allows applications to stay mostly untouched, while replacing machinery behind the scenes - feed handlers, ticker plant and transport distribution, for example. So, data can be delivered without clients needing to change a legacy API to connect to the application, nor is it necessary to change the application. The point is that latency is reduced and costs are kept low.

Investment banks may be willing to pay for bleeding edge latency, but some asset managers might not. Gil Tene, CTO at Azul Systems, said that for some kinds of clients, time to market is a critical factor. "Being fast but taking 10 years to get to the market is useless," he said. "And there is an obvious cost part of this equation -how many people and how much manpower are you going to throw at the problem?"

From his vantage point, Tene sees Java being heavily adopted in low latency systems in spite of its reputation for inconsistency because it reduces both cost and time to market. In other words, there is a programming productivity benefit with the use of Java.

Along the way, Azul says it is able to also get consistent execution - an area it specialises in. "By eliminating the natural inconsistencies in the platform itself - like garbage collection - we take away a tremendous amount of engineering people have to do to work around those limitations in order to gain the other benefits," he said.

Since 2007, the core density of Intel's Xeon-based platform has increased six times. Such a gain has led to changes in the ways trading applications perform and are deployed. Intel's "abundance of cores", Tene said, has made it possible to take technology to commodity servers, and into very latency sensitive markets.

On the hardware side, IBM's John Encizo, senior technical sales specialist, said that Intel's chip development has also led to a large amount of differentiation that can present an obstacle when disparate teams need to work together.

"There is a hardware engineering and an application or software engineering group. It's actually getting the two teams to communicate to understand the differences in the platforms and that each platform does have its unique benefits, features, and options," he said.

A recording of the webinar can be accessed HERE.