Date:   Wed, 26 Jul 2023 21:32:33 +0200
From:   Pratyush Yadav <ptyadav@...zon.de>
To:     Keith Busch <kbusch@...nel.org>
CC:     Christoph Hellwig <hch@....de>, Sagi Grimberg <sagi@...mberg.me>,
        "Jens Axboe" <axboe@...nel.dk>, <linux-nvme@...ts.infradead.org>,
        <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] nvme-pci: do not set the NUMA node of device if it has
 none

On Wed, Jul 26 2023, Keith Busch wrote:

> On Wed, Jul 26, 2023 at 05:30:33PM +0200, Pratyush Yadav wrote:
>> On Wed, Jul 26 2023, Christoph Hellwig wrote:
>> > On Wed, Jul 26, 2023 at 10:58:36AM +0300, Sagi Grimberg wrote:
>> >>>> For example, AWS EC2's i3.16xlarge instance does not expose NUMA
>> >>>> information for the NVMe devices. This means all NVMe devices have
>> >>>> NUMA_NO_NODE by default. Without this patch, random 4k read performance
>> >>>> measured via fio on CPUs from node 1 (around 165k IOPS) is almost 50%
>> >>>> less than CPUs from node 0 (around 315k IOPS). With this patch, CPUs on
>> >>>> both nodes get similar performance (around 315k IOPS).
>> >>>
>> >>> irqbalance doesn't work with this driver though: the interrupts are
>> >>> managed by the kernel. Is there some other reason to explain the perf
>> >>> difference?
>>
>> Hmm, I did not know that. I have not gone and looked at the code but I
>> think the same reasoning should hold, just with s/irqbalance/kernel. If
>> the kernel IRQ balancer sees the device is on node 0, it would deliver
>> its interrupts to CPUs on node 0.
>>
>> In my tests I can see that the interrupts for NVME queues are sent only
>> to CPUs from node 0 without this patch. With this patch CPUs from both
>> nodes get the interrupts.
>
> Could you send the output of:
>
>   numactl --hardware

$ numactl --hardware
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
node 0 size: 245847 MB
node 0 free: 245211 MB
node 1 cpus: 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63
node 1 size: 245932 MB
node 1 free: 245328 MB
node distances:
node   0   1
  0:  10  21
  1:  21  10

>
> and then with and without your patch:
>
>   for i in $(cat /proc/interrupts | grep nvme0 | sed "s/^ *//g" | cut -d":" -f 1); do \
>     cat /proc/irq/$i/{smp,effective}_affinity_list; \
>   done

Without my patch:

    $   for i in $(cat /proc/interrupts | grep nvme0 | sed "s/^ *//g" | cut -d":" -f 1); do \
    >     cat /proc/irq/$i/{smp,effective}_affinity_list; \
    >   done
    40
    40
    33
    33
    44
    44
    9
    9
    32
    32
    2
    2
    6
    6
    11
    11
    1
    1
    35
    35
    39
    39
    13
    13
    42
    42
    46
    46
    41
    41
    46
    46
    15
    15
    5
    5
    43
    43
    0
    0
    14
    14
    8
    8
    12
    12
    7
    7
    10
    10
    47
    47
    38
    38
    36
    36
    3
    3
    34
    34
    45
    45
    5
    5

With my patch:

    $   for i in $(cat /proc/interrupts | grep nvme0 | sed "s/^ *//g" | cut -d":" -f 1); do \
    >     cat /proc/irq/$i/{smp,effective}_affinity_list; \
    >   done
    9
    9
    15
    15
    5
    5
    23
    23
    38
    38
    52
    52
    21
    21
    36
    36
    13
    13
    56
    56
    44
    44
    42
    42
    31
    31
    48
    48
    5
    5
    3
    3
    1
    1
    11
    11
    28
    28
    18
    18
    34
    34
    29
    29
    58
    58
    46
    46
    54
    54
    59
    59
    32
    32
    7
    7
    56
    56
    62
    62
    49
    49
    57
    57

-- 
Regards,
Pratyush Yadav



Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Jonathan Weiss
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879

Powered by blists - more mailing lists