[RFC PATCH 0/5] Enable use of Solid State Hybrid Drives

[RFC PATCH 0/5] Enable use of Solid State Hybrid Drives

[Dan Williams]

On Wed, Oct 29, 2014 at 8:28 PM, Martin K. Petersen
<martin.petersen@oracle.com> wrote:
> The next step was trying to map these hints into what was available in
> xadvise(), NFS 4.2 and the recent T10/T13 efforts. That wasn't trivial
> and there really isn't a 1:1 mapping that works. So I went to T10 and
> tried to nudge things in the same direction as NFS 4.2. Mainly because
> that's closer to what we already have in xadvise().

In case you still have hair left to pull wrangling these multiple
specifications, Matthew reminds me that NVME also has cache advice at
the transport layer.

> Jens> I think we've needed a proper API for passing in appropriate hints
> Jens> on a per-io basis for a LONG time.
>
> Yup.

I understand the desire to have per-io / per-inode xadvise()-style
hints, but I don't see why not also include a per-pid capability?

Per-pid was not "icky" for flashcache [1].  It let's you flag
processes that should not pollute the cache, as well "cache warming"
processes pre-loading sub-ranges of files that is awkward to do with a
per-inode hint.  Per-pid also allows hinting on behalf of other
otherwise cache-unaware processes.

> Jens> That is the big challenge. We've tried (and failed) in the past to
> Jens> define a set of hints that make sense. It'd be a shame to add
> Jens> something that's specific to a given transport/technology.
>
> Absolutely!

In this RFC we end up punting the ultimate kernel-to-transport hint
translation to userspace.  The kernel has a default interpretation,
but it seems it will almost always be inadequate trying to account for
per-device-quirks and platform performance policies.

[1]: https://github.com/facebook/flashcache/blob/master/doc/flashcache-doc.txt#L139

[RFC PATCH 0/5] Enable use of Solid State Hybrid Drives

[Jens Axboe]

On 2014-10-29 22:19, Dan Williams wrote:
> I understand the desire to have per-io / per-inode xadvise()-style
> hints, but I don't see why not also include a per-pid capability?
>
> Per-pid was not "icky" for flashcache [1].  It let's you flag
> processes that should not pollute the cache, as well "cache warming"
> processes pre-loading sub-ranges of files that is awkward to do with a
> per-inode hint.  Per-pid also allows hinting on behalf of other
> otherwise cache-unaware processes.

per-pid is imho fine as well, as long as it's not the primary interface. 
I quite like how the io priority works in this regard. If the task has a 
priority set, we use that. If you pass in something else, that overrides 
the task set one.

per-pid allows you to modify how we treat applications without modifying 
the application itself. This is handy for eg streamed backup and 
similar, which is most likely why flashcache has it.

-- 
Jens Axboe