Azul Works Low-Latency Space06.18.2013
While some professional programmers in the low-latency space may eschew Java, Azul Systems Chief Technology Officer Gil Tene says demand for the 18-year-old programming language is underpinned by the beneficial tradeoffs of cost, productivity, the talent pool for algorithmic programmers and the time to market.
A strong trend towards Java emerged in the past five or six years, and some of the world’s biggest banks, hedge funds, proprietary traders, online retailers and other clients are using the highly elastic, Java runtime platform now, Tene told Markets Media last week on the sidelines of an algorithm conference in Chicago. OptionsCity Software is an example of a ‘100% Java shop’.
The introduction of Azul’s Zing Java Virtual Machine (JVM) for Java-compatible high performance, low latency, and ability to support large in-memory data sets without performance penalty has helped restore some luster to Java as a code language. Zing “takes care of the Java hiccups,” countering the risk of stalls that had engendered a reputation for inconsistent execution, its creators said.
From exchanges’ point of view, with hiccups rarer and dramatically smaller, market data flows more smoothly, Tene said.
Unlike other commercially available JVMs, Zing makes use of highly optimized compilers as well as Azul’s Continuously Concurrent Compacting Collector (C4) technology, implementing a pause-less garbage collection algorithm. This allows enterprise developers to create and deploy new, in-memory applications without risking the problematic performance artifacts caused by traditional Java garbage collectors, whose role is to free up space.
OptionsCity has integrated Zing into Freeway, its server-based, multi-asset algorithmic trading platform, and Metro, its fully configurable, high-performance market-making and electronic-trading platform.
In April, LMAX Exchange – the first multi-lateral trading facility for foreign exchange regulated by the U.K.’s Financial Services Authority – partnered with Azul to put Zing under the hood for continuous performance optimization of its low-latency execution venue.
“Zing eliminates operational interruptions, response time outliers and the scalability limits of Java-based trading and other financial services applications,” LMAX said in a release, adding that it would not need to make any changes to its trading software architecture, approach to software development, or advanced continuous delivery process.
“Even under demanding and unpredictable loads, Zing enables financial institutions like LMAX Exchange to seamlessly grow transaction volumes and meet ever-stricter SLAs (Service Level Agreements),” added Azul Systems chief executive Scott Sellers.
Looking ahead, Azul Zing is invested in specific features for low-latency trading and is tackling areas like the warm-up problem. Tene said he counts how many people are experiencing the same problems and focuses on solutions to commonly shared issues.
Speeding orders into the market immediately at the open is critical to some firms, and some algos take time to get warmed up for the first transaction of the session. Azul recommends how customers can tweak their code, or how to work with their system so it performs better for their needs, and is adding product features that will provide market opening reliability in response to this concern. Sellers and Tene combined hold more than 40 patents on computing technologies.
“Speed and scale only matter if you meet the consistency requirements,” Tene said.