With the new Options Price Reporting Authority (OPRA) feed set to go live on 9 October, Exegy has developed a field-programmable gate array-based (FPGA) managed appliance which it claims can handle the anticipated deluge of data. Exegy CEO David Taylor spoke to BEST EXECUTION about the firm’s latest piece of hardware, and how they have future-proofed the technology.
By the end of 2022, US stock options surpassed 10 billion contracts, with single-stock and index options volumes more than doubling since 2019. Now, OPRA’s guidance states that participants should prepare for capacity of 120+ million messages per second following the expansion.
Exegy says a single one of its servers handling the OPRA feed could reduce rack space requirements in data centres by as much as 15 to 20 times compared to a software-only solution.
Speaking to BEST EXECUTION, Exegy CEO David Taylor explained how the changes in posted order prices and resulting Best Bid and Offer (BBO) quotes represent the largest portion of market data message volume.
“In the US equities markets, it is common for there to be more than 100 quote messages for every trade message. Thus, if 10x more trades occur in a stock, resulting in higher volatility, the change in market data message volume can be 100x larger.”
The issue is compounded in the US options market where there can be more than 100,000 option contracts for a highly liquid stock. “Every time the price of the stock changes, the values (and therefore prices) of all of the option contracts change. This is why we see more than 15,000 quote messages for every trade message in the US options market data feeds, as well as feed volumes that are 10x to 100x higher relative to the equity market feeds.”
This increasing options market data is combined with the need to process this with consistently low latencies.
In response to these ballooning volumes, an FPGA-based solution offers huge advantages, Exegy says, in terms of both scalability and higher throughput. FPGAs allow the firm to design purpose-built circuits for processing market data that are faster and more efficient than general purpose processors running software.
“Using a manufacturing analogy, an FPGA allows us to build multiple, parallel assembly lines where each step in each assembly line is specifically designed to perform a market data processing step as quickly as possible. We can build 10s or 100s of parallel assembly lines with many stages in each assembly line.”
By contrast, a general purpose processor running software is akin to having a highly skilled artisan who reads and executes instructions on how to build widgets very quickly.
“It is easy to see how the FPGA approach can yield 10x to 1000x higher throughput than the general purpose approach,” Taylor said.
With the number of contracts and data volumes set to only increase, Taylor said there are two key elements to delivering future-proofed solutions for customers.
“First, we design our appliance platforms to have excess processing capacity and to allow for remote configurability of the FPGA logic and software. Second, we provide a managed service that packages and delivers our continuous innovation in the form of firmware and software updates to deployed appliances that increase capacity and reduce latency.” Taylor said.
“It’s like buying a car that gets 2x faster with 4x better fuel economy over your three to five year ownership period.”
©Markets Media Europe 2023
