Very Promising Linux Patch Optimizes TLB Flushes During Page Reclamation

[phoronix]

Phoronix: Very Promising Linux Patch Optimizes TLB Flushes During Page Reclamation

Google engineer Vinay Banakar sent out a patch this week for the Linux kernel's memory management code to optimize TLB flushes during page reclaim and are showing very promising results...

https://www.phoronix.com/news/Linux-Optimize-TLB-Flush-Reclaim

[Kjell]

This reminds me of enabling transparent_hugepage=madvise to improve application start up time

https://www.reddit.com/r/archlinux/comments/1atueo0/higher_ram_usage_since_kernel_67_and_the_solution/

[LtdJorge]

Small change but potentially huge. On systems with high memory pressure under normal circumstances, like CI workloads, this should improve performance quite a bit.

See Meta's oomd, created because their CI servers were consistently crashing due to oom-killer shooting down unrelated processes. In such high memory pressure scenarios, oomd will have to work a lot, before the oom-killer or oomd have to make a decision. So the overall impact described in the MR will be high for those servers, which I can only assume are highly threaded.

[damentz]

Small change but potentially huge. On systems with high memory pressure under normal circumstances, like CI workloads, this should improve performance quite a bit.

See Meta's oomd, created because their CI servers were consistently crashing due to oom-killer shooting down unrelated processes. In such high memory pressure scenarios, oomd will have to work a lot, before the oom-killer or oomd have to make a decision. So the overall impact described in the MR will be high for those servers, which I can only assume are highly threaded.

Yes, from what Vinay described, most of the problems with swap out aren't actually the swapping or memory reclaim, it's the IPIs to all cores for every single page that causes massive throughput reduction for running applications. This is really exciting and might be massive on low memory systems.