STAC Confirms Exablaze As The Fastest Tick-To-Trade System With Its New FDK Extension Pack - Exablaze Slashes Latency By More Than Half In Independent Tests

Leading global provider of ultra-low latency network devices, Exablaze, has broken the previous best STAC-T0™ benchmark results, with its newly-announced ExaNIC FDK-XP (Firmware Development Kit Extension Pack). It slashed latency to less than half the previous benchmarked best, recording actionable latencies¹ of 31-44 nanoseconds, signalling a marked advance in tick-to-trade system performance.

The independent benchmark results underscore the superiority of the company’s devices in bringing the lowest-possible latency between market data arrival in trading platforms (‘ticks’) and the moment a ‘trade’ message is sent, a crucial pursuit for firms engaging in low latency trading.

In the STAC benchmarks, Exablaze demonstrated a high degree of determinism, with maximum latency remaining constant even while running at 11 million packets per second (7.72 Gbps), 3.4 times the maximum ingress (message) rate tested previously². Exablaze set new records in every measurement.

Exablaze was evaluated using the STAC-T0 benchmark³, developed by the Securities Technology Analysis Center (STAC®), the independent testing agency that facilitates the STAC Benchmark Council, an industry group consisting of financial firms and technology vendors.

According to STAC, "The T0 benchmark tests tick-to-trade network latency of any trading platform to a high degree of accuracy, regardless of the hardware or software being used”. In this case, Exablaze supplied a system built using its new extended firmware development kit (FDK-XP). Because of the exceptionally high performance of Exablaze, the benchmark had to be adjusted to measure latency in picoseconds (1 trillionth of a second), rather than nanoseconds, as it did previously. Importantly, the benchmarking suite was unable to exceed Exablaze’s bandwidth capabilities.

Dr. Matthew Grosvenor, Vice President of Technology at Exablaze, commented: “We’re absolutely thrilled to be able to announce these record-breaking results. Reducing latency by more than 50% will give trading firms and teams a significant advantage. We’re also very proud of the high bandwidth and determinism at which we’re doing this. Processing trades with ultra-low latency is difficult, but doing it at high bandwidth is exceptionally hard. Our clients can be confident that even with very high market volumes our system will maintain its performance.”

Grosvenor added: “It’s great to see that our new FDK Extension Pack is performing even better than we expected. One of the unique aspects of these benchmark results is that our solution uses only Exablaze components and is therefore fully-integrated, allowing for a much higher rate of performance than any other system available. We offer all the components for traders to build (and measure) their own trading systems.”

The FDK-XP was a key factor in this astonishing performance. A common problem when building low latency tick-to-trade systems (like the one required by STAC-T0) is connecting the FPGA decision-making logic to TCP networking endpoints. FDK-XP addresses this problem by providing full TCP transmission support in FPGA firmware while keeping latency to an absolute minimum. Its two main components are an ultra-low latency PCS/MAC for connecting FPGA logic to Ethernet networks, and a hybrid software/hardware accelerated TCP/IP Engine (ATE) for connecting FPGA logic to TCP/IP endpoints.

Notes

The full STAC Report™ is available at www.STACresearch.com/EXA181106

STAC-T0 defines actionable latency as the difference between the moment a system has received all the information necessary to trigger a trade and the moment that it begins sending the trade order.
The benchmarks were carried out using Exablaze components end-to-end: the ExaNIC V5P Network Application Card, ExaNIC Firmware Development Kit Plus (FDK-XP), ExaSOCK socket acceleration system and ExaSOCK ATE (Accelerated Transmit Engine) for hardware accelerated TCP. Only a basic server was used to run the trading system, in this case a low-end Dell R230. Exablaze components were also verified by STAC and used for the packet timestamping and capture aspects of the tests.
STAC-T0 benchmark was defined and specified by a number of banks and hedge funds in the STAC Benchmark Council. Its tests are designed to measure the performance of processing tick-by-tick market data, i.e. a ‘tick database’, with well-defined hardware and software technology stacks.