NovaSparks has set up a partnership with Thesys Technologies to provide a low-latency market data service to Thesys customers. The service is based on NovaSpark’s field programmable gate array (FPGA) hardware architecture, including a feed handler that allows efficient and cost effective access to market data with minimal impact on the resiliency and latency of trading systems.
Luc Burgun, president and CEO of NovaSparks, says that compared to high frequency trading firms and investment banks establishing their own systems, the NovaSparks solution delivers cost savings while providing low-latency access to market data. This is achieved using FPGA technology that replaces a number of steps in the chain, such as the switches and processors that normally run feed handlers in software, and allows access to the order book in around one microsecond.
Burgun explains: “It might be more cost effective to run software, but it is much slower because you have a lot of latency. In particular, when the market is very fast and there is a lot of activity, software can really slow you down. It can take five to ten microseconds to provide data to the trading software.”
The low-latency market data service allows each customer to have its own list of symbols and conduct calculations while sharing the FPGA, reducing overall cost and rack space required by trading firms at colocation centres. Burgun acknowledges competitors in the market offering FPGA services, but says NovaSpark’s coverage of over 30 markets and its format normalisation make its service particularly valuable for the efficient delivery of market data.
Looking forward, Burgun says NovaSparks’s goal is to expand coverage. With over 30 stock markets covered at the moment, he explains that the focus in future will be on cash equities, futures and options, with a view to consolidating market coverage across the globe. He concludes: “We have a lot of things coming up in terms of improving our market coverage and providing fast services based on FPGAs. As usual, we are going to work on improving latency and, ultimately, we are looking to reduce the time the market data is in the FPGA to the time it is available in the application programming interface of the trading software to below one microsecond.”