The Next-Gen Ticker Plant
To compete with HFT firms, a Tier-1 Global Investment Bank needed to overcome the operating system kernel limits slowing down their C++ ticker plant. The engineering team evaluated a move to Field Programmable Gate Arrays (FPGA) to process data directly on the network card, achieving a 100x reduction in latency by burning the logic into hardware and bypassing the OS entirely.
CONTEXT (THE BLOCKAGE)
A [TIER-1 FINANCIAL] needed to reduce the latency of their market data feed to compete with HFT firms, but the C++ stack was hitting the limits of the OS kernel.
THE ACTION
Evaluated a move to FPGA (Field Programmable Gate Arrays) to process market data directly on the network card, bypassing the OS entirely.
THE RESULT
Validated a 100x reduction in latency, moving from microseconds to nanoseconds.
In HFT, the Operating System is the enemy. The ultimate optimization is to remove the software and burn the logic into hardware
[Ref: TI-010]