Linux 6.7 To Boast Better Performance For FQ Packet Scheduling Algorithm

The Linux kernel's Fair Queue "FQ" network packet scheduling algorithm that is celebrating its 10th anniversary since being mainlined in the Linux 3.12 kernel cycle is celebrating by rolling out some performance optimizations with the next version of the Linux kernel.

Eric Dumazet of Google has been working on a round of performance improvements to the Fair Queue network scheduling algorithm code. The end result is seeing around a 5% throughput improvement and up to a 13% increase for unpaced UDP packets.

"net_sched: sch_fq: round of improvements

For FQ tenth anniversary, it was time for making it faster.

The FQ part (as in Fair Queue) is rather expensive, because we have to classify packets and store them in a per-flow structure, and add this per-flow structure in a hash table. Then the RR lists also add cache line misses.

Most fq qdisc are almost idle. Trying to share NIC bandwidth has no benefits, thus the qdisc could behave like a FIFO.

This series brings a 5 % throughput increase in intensive tcp_rr workload, and 13 % increase for (unpaced) UDP packets."

These FQ packet scheduling improvements have been merged into the net-next branch ahead of the Linux 6.7 kernel merge window opening up in about one month's time.