PMFS and Linux 4.0 DAX

[Jack Harvard]

Hi Matthew,

I'm just trying to understand more, not trying to disagree but trying
to be convinced. As you said there's no big win to be had by enabling
byte access to filesystem metadata that are stored on the persistent
memory, presumably you have measured such across a range of workloads?
including file utilities such as mv, rename, touch etc which do only
meta data manipulations as mentioned in the PMFS paper.

I suppose the persistency support still needs to be built on top of
the DAX, or does the file APIs like read() and write() have been
rewritten with such, which would need to support different
architectures if ext4 already worked with a few architectures.

Thanks

Jack
Jack Harvard


On Thu, Apr 23, 2015 at 3:37 PM, Matthew Wilcox <willy at linux.intel.com> wrote:
> On Thu, Apr 23, 2015 at 11:50:19AM +0100, Jack Harvard wrote:
>> The recently released Linux 4.0 ext4 filesystem can be enabled with
>> DAX which bypasses the page caches, looks it's not byte-addressable,
>> but still using a block device/ram driver for accessing the PM.
>>
>> Wondering whether there is a list of differences between ext4 with DAX and PMFS?
>
> PMFS is buggy and unsuitable for use in a production environment.
> We looked at fixing it, and decided it was unfixable without a rewrite.
> DAX is the approach we decided to pursue instead.
>
> The fact that it sits on top of a block device isn't terribly relevant ...
> user data is read/written through byte accesses.  Filesystem metadata
> still goes through the block I/O path, but we don't think there's a big
> win to be had by converting the filesystem to do byte accesses instead.