Mon, 11 Jun 2012 02:33:50 GMT - NovaSparks
The first company to put all market data processes into an FPGA platform to achieve deterministic nanosecond speeds for processing the feed, building and managing the Order Book
NovaSparks, an FPGA market data solution company, launches a pure FPGA Order Book Builder capability. The Order Book building function uses the NovaSparks FPGA Market Data Matrix™ architecture and provides deterministic nanosecond speeds. NovaSparks is the first company to commercialize a pure FPGA appliance architecture for market data processing.
The entire data processing of the feed including the Order Book building achieves deterministic processing latency of 880 nanoseconds (ns) on average, under 1 microsecond 99% of the time and between 1 microsecond and 1.5 microsecond 1% of the time. Unlike CPU architectures or hybrid approaches that use a CPU with an FPGA card, this pure FPGA architecture achieves constant, deterministic latency regardless of data rates, the number of venues and symbols, or the number of downstream users.
The FPGA Book Builder transmits a normalized stream of updates for the trading algo to use. The depth of book is configurable by the user. The Book Builder outputs a single format, called a normalized stream, regardless of the data format and protocol used by the venue. This normalized output enables trading teams to develop algos for one venue then scale the algo across multiple venues without changing the core algo logic.
“We believe there is no faster view of the market because no one else has put all market data processing functions together into an FPGA architecture,” explains Yves Charles, CEO of NovaSparks. “This announcement proves the FPGA Market Data Matrix architecture can tackle more complex tasks and can scale to meet the ever-increasing performance needs of the market,” suggests Charles.
Senior Director of Research for Aite Group, Adam Honoré comments “The performance gap between pure FPGA architectures and hybrid approaches is widening. As more features become available through hardware acceleration, Trading Firms will have to routinely evaluate whether or not their solution remains competitive in the market.”
SIFMA Booth #1502: See a Live Demos of pure FPGA Order Book with Packet-by-Packet Latency Tracking
|
Market Data Feed Handler Functions
|
NovaSparks Market Data Matrix
conducts all functions in FPGA
|
|
A/B Feed Arbitration
|
FPGA
|
|
Parsing/Decoding
|
FPGA
|
|
Instrument Filtering
|
FPGA
|
|
Book Building/Management
|
FPGA
|
|
Normalization
|
FPGA
|
|
Distribution (PCIe and 10Gig)
|
FPGA
|
|
Algo
|
FPGA or CPU
|
Background:
In most Equity Cash markets, the order book is not pre-formatted for the algo; the exchange transmits each individual order and it is the responsibility of the feed handler to collect all packets, decode the messages, assemble and maintain the depth of prices and volume to create ‘the book’ of supply and demand for a given symbol.
This task is traditionally done in software using a CPU-based architecture. However the CPU-based architectures exhibit queuing behaviors, which slows down the processing in volatile or bursty markets. NovaSparks eliminates the CPU and uses a matrix of FPGAs to process and normalize market data.
About NovaSparks www.novasparks.com
Contact: Marie Giangrande, Marie@publicnotions.com +1.203.536.1334
NovaSparks, inventor of the FPGA Market Data Matrix™ architecture, is an engineering company dedicated to FPGA solutions for Capital Markets. Taking a unique ‘FPGA- centric’ design approach, NovaSparks is the first company to design and commercialize a set of pure FPGA Feed Handlers offering deterministic, nanosecond processing speeds. Available for major Equity and Futures venues across Europe and North America, the Feed Handlers are packaged in ‘easy-to-deploy’ appliances and can be chained together to seamlessly scale.
Founded in 2008 and backed by Partech International, NovaSparks’ executive management team is based in Paris France, Boston MA USA, London and New York. www.novasparks.com
