NovaSparks Slashes Feed-Handler Latency by 25%


BUSINESS WIRE – NEW YORK & LONDON & PARIS–NovaSparks™, the leading field programmable gate array (FPGA) market data company, today announced it has completely redesigned its book building engine inside NovaTick, its latest generation ticker plant, dramatically lowering the processing latency for order based feeds. The new and improved book building engine is at the heart of NovaSparks’ pure FPGA architecture. One of the other architectural changes includes a new ultra-low latency PCS (Physical Coding Sublayer). This improvement is especially valuable to NovaLink customers, who use the FPGA-to-FPGA link to receive NovaSparks’ ticker plant data directly into their own FPGA trading systems. These architectural changes will be formally available for the Nasdaq TotalView ITCH feed in November, followed by all the other order based feeds currently supported by NovaTick in January.

“It has been one year in the making; we managed to reduce latency by 300ns and it will benefit all of our customers, whether they consume the market data in 10Gig Ethernet, PCIe or NovaLink,” said NovaSparks CEO Luc Burgun. “This new book building architecture also puts us on very solid ground to dramatically increase the scalability of our solutions in the near future well beyond the currently supported 8,191 instruments for order based markets. This consolidates our position as the leading pure FPGA market data company.”

NovaTick’s design is based on a pure FPGA architecture where all real-time market data processing functions are performed in hardware resulting in the lowest and most deterministic latency. This FPGA-centric architecture is augmented with tightly integrated multi-core Intel processors to support non-latency sensitive tasks such as exchange recovery functions. The redesigned book building engine benefits all the output interface options of NovaTick as shown in the table below:

Output Option
Processing latency for the NASDAQ TotalView ITCH feed
configured for the full universe of instruments.
Current Architecture New Architecture
10Gig Ethernet Multicast
measured wire-to-wire
1100 nanoseconds 800 nanoseconds
measured wire-to-trading-FPGA-core
1050 nanoseconds 750 nanoseconds
PCIe Direct Memory Access
measured wire-to-server-memory
1550 nanoseconds 1250 nanoseconds

Output interface options of NovaTick, NovaSparks’ latest generation platform, and their resulting latency.

NovaLink is the FPGA-to-FPGA link that enables trading firms to receive NovaSparks’ ticker plant data directly into their own FPGA trading systems. The data is consumed through a normalized hardware application program interface (API) provided by NovaSparks as an intellectual property (IP) core. By using NovaLink, trading firms off-load the feed handler functions of their overall design onto separate FPGA chips, which gives more space on their own FPGA to implement larger scale trading strategies that will not fit on a single chip system otherwise.

NovaSparks will be exhibiting at the FIA Expo in Chicago on October 19 and 20, 2016. Visit NovaSparks at booth #713 in the blue hall.

For more information about the new book building engine, NovaTick, or NovaLink, please contact sales@novasparks.com.

Related articles

  1. OPINION: Artificial, Yes. Intelligent? Maybe.

    The platform uses machine learning and AI to automate fixed income portfolio construction and management.

  2. Buy Side Forced to Review Collateral Arrangements

    As rates and volatility rise, the need to optimise the value of collateral is greater than ever.

  3. This opens the door for the delivery of the first live intraday repo market.

  4. Cloud-enabled infrastructure delivers a 10% performance improvement for participants.

  5. Technologists can accelerate delivering software to power trading, risk management, and compliance.