[RFC PATCH 0/6] nvme multipath eBPF path selector
Mike Christie
michael.christie at oracle.com
Tue Jul 29 07:53:09 PDT 2025
On 7/29/25 2:54 AM, Christoph Hellwig wrote:
> On Tue, Jul 29, 2025 at 09:06:47AM +0200, hare at kernel.org wrote:
>> From: Hannes Reinecke <hare at kernel.org>
>>
>> Hi all,
>>
>> there are discussion on having to deploy more complex I/O scheduling
>> algorithms for NVMe,
>
> Are "we"? Where?
I think that's me. I was originally looking into eBPF path selectors
for ADNN:
https://www.snia.org/educational-library/adaptive-distributed-nvme-namespaces-2020
I was bugging Hannes about what happened to the spec. The idea was
just that if the spec is dead, then something like eBPF path selectors
could be an option.
I'm not currently looking at it, because I got busy with other
work. But, maybe the ceph people would want it for their nvme frontend
like in the video. For the ceph iscsi project, I had looked at SCSI
referrals, and it didn't fit (I can't remember the details), and ADNN
looked promising.
>
>> but then there's the question whether we really
>> want to carry these in the kernel.
>
> If it makes sense, "we" do, yes.
>
> "We" don't want to export random crap for sure.
>
>
More information about the Linux-nvme
mailing list