About this question
Hard · Networking & Systems · Quant Trader interview question · kernel, ring-buffer, io_uring, performance, latency
A high-frequency trading firm is optimizing its market data ingestion pipeline. They are considering using a kernel ring buffer (like io_uring) instead of traditional system calls for reading incoming network packets. Assume that switching to a kernel ring buffer eliminates 80% of the overhead associated with each system call. Furthermore, the system currently makes 10,000 system calls per second. If the typical system call overhead takes 5 microseconds, and all other processing remains constan