CME was one of the first exchanges that we supported. We were there when CME moved from the FAST-encoded MDP 2 to the improved and faster MDP 3 with SBE (Simple Binary Encoding) around seven years ago. That’s when we released our original MDP 3 feed handler.
Since then, we have added support for various other exchanges in North America, Europe, and Asia. We have done many customer projects and gained a lot of experience.
We’ve managed to reduce the latency by 30-100 nanoseconds, depending on the specific use case and hardware platform.
Today, we are taking that experience back to CME, as we release a completely reworked CME MDP 3 feed handler. We’ve managed to reduce the latency by 30-100 nanoseconds, depending on the specific use case and hardware platform. The total wire-to-wire latency gets below 300 nanoseconds with a full tick-to-trade pipeline, including the book, in FPGA. For simpler strategies triggering directly on market data messages, the latency is close to 100 nanoseconds.
The new CME MDP 3 feed handler is faster, leaner, and better. We really like the new code. That being said, we can’t wait to rework it again a couple of years from now in our endless pursuit for improvement.
What do you think? Are you ready to switch gears now?