| 1 |
On Fri, Sep 24, 2021 at 10:35 PM William Kenworthy <billk@×××××××××.au> |
| 2 |
wrote: |
| 3 |
|
| 4 |
> In going down the NUMA rabbit hole, I discovered "irqbalance". Does |
| 5 |
> anyone have an opinion on its usefulness? It is in portage. |
| 6 |
> |
| 7 |
> On some multicore arm systems I am using irq affiity to steer certain |
| 8 |
> irq's to faster CPU's (network, usb) - but from what I have been reading, |
| 9 |
> irqbalance can improve a mixed workload but a system with a small number of |
| 10 |
> busy irq's is better served by separating and locking them to different, |
| 11 |
> more powerful processors. e.g., arm big.LITTLE architectures. |
| 12 |
> |
| 13 |
IIRC MSIs have largely addressed the issue, so irqbalance is not so useful |
| 14 |
anymore. Eg, /proc/interrupts on this system shows that the nvme drive gets |
| 15 |
32 interrupts, and the intel gig eth card gets eight interrupts per port, |
| 16 |
so there's no busy interrupts |
Archives