Metamako, the leading specialist provider of high-performance networking platforms, has launched a new series of devices capable of running up to 3 powerful FPGAs on a single platform.
These are the most powerful FPGA-based devices in their class, designed to provide financial institutions and vendors with enormous processing power, which can be connected directly to exchanges.
This configuration allows multiple performance-sensitive applications, such as trading algorithms, pre-trade risk checks or market data processing to run on a single device. Trading firms may programme all three FPGAs to operate in concert, bringing the combined power to bear on a single trading strategy.
The new range of devices leverages the latest FPGAs from Xilinx, the Kintex® UltraScale™ and the Virtex® UltraScale+™. Clients deploying the triple FPGA solution will benefit from near-zero latency links between each FPGA, and will see significantly reduced latency, compared with FPGAs connected via serial links.
Dave Snowdon, CTO of Metamako, says: “Having such a large amount of power so close to the network edge will revolutionise low latency and determinism in trading. To truly win the low-latency race, traders must connect directly to the exchange, using the shortest fibres possible. That means getting as much FPGA fabric as close to the exchange as possible.”
He continues: “Being able to offer the option of up to three large FPGAs marks our move into a significantly more powerful and faster FPGA-enabled networking and applications platform. Our clients and ecosystem partners will now have much more flexibility; for example, they will be able to scale up the number of trading apps, and run larger and much more complex apps. This will also reduce latency and allow ultra-fast, direct communication to exchanges. Combined with the capabilities of these cutting-edge FPGAs, application developers can avoid off-chip RAM, providing a leap in the performance of their FPGA-based applications.”
Metamako’s ground-breaking approach provides a sophisticated solution, with lower latency. It is also much simpler and considerably more convenient to implement compared to the traditional approach of using FPGAs embedded on PCI Express cards, thanks to having multiple components embedded in a single device.
All core Metamako features such as market data fan-out, scalable broadcast, connection monitoring, patching,tapping, time-stamping and deterministic port-to-FPGA latency of just 3 nanoseconds, can be leveraged - including between the FPGAs.