Matt Benjamin

@leseb I was having trouble running our own test suite locally, so decided to try with the go-ceph code

> Hi all, it's been a while since this PR saw any activity. Based on previous comments I assume @mattbenjamin is not interested in driving this to completion. @leseb are...

sorry, I'll try to hot this up

hi @liyinshubyte ; what you describe in your second diagram is closer to what cohortfs did with ceph pnfs: we returned a layout to the client that allowed the client...

@liyinshubyte in the first instance, you need to decide what set of rules your implementation will follow. cohortfs experimented with at least 2 different *layouts*. First, we experimented with the...

(As @dang explained earlier, our most advanced prototypes were based on our own programmable layout. That kind of approach would require you to put custom logic in to NFSv4 clients--which...

The idea of going direct to the OSDs was to remove hops/proxying. It relied on the ability of Ceph or a similar scale-out storage at sub-file granularity to provide improved...

> For pNFS to work, the NFS client needs network access to the Data Servers (DSs) in addition to the MetaData Server (MDS). In normal NFS, a single Ganesha would...

> I'm not sure that Ganesha currently really has usable pNFS. Some FSALs are set up for it (GPFS for sure, Gluster? Others?) but I don't know of anyone who...

> I've researched pNfs for a few days , kernel-Nfsv41 does not contain pNfs. Then I found ganesha which declare pNfs support . > > If you discourage user to...