Price your future right. 60% off with code BLACKSCHOLES at USD checkout. Offer ends in 8d 09h 54m 23s8 days 9 hours 54 minutes 23 seconds.
Price your future right. 60% off with code BLACKSCHOLES at USD checkout. Offer ends in 8d 09h 54m 23s8 days 9 hours 54 minutes 23 seconds.

Sign in to watch this lesson
Trading Networks · 4:03
We follow one packet up the Linux receive path, from the NIC's DMA into an RX ring through the interrupt, the softirq and NAPI poll, IP and UDP, the socket queue and the copy in recvfrom(). Then we time the software half on this Mac's loopback: about 18 µs to reach the socket, about 10 µs more to wake a sleeping receiver, cut to about 2.5 µs by spinning. Using OSTEP's interrupts-versus-polling trade-off, we show why trading firms bypass the kernel with Onload/ef_vi, DPDK or RDMA, and what that costs them (industry practice, not K&R).