On 08/11/2017 09:21 AM, Frank Filz wrote:
That seems overkill to me. How many strategies would we support (and
test)?
Part of the problem is that we've drastically changed how FDs are handled.
We need to rethink how LRU should work in that context, I think.
I wonder also if taking pinning out of the equation (which moved cache
objects that had persistent state on them into an entirely separate queue)
has had an effect.
Could be.
Hopefully those objects get quickly promoted to MRU of L1
(since they should have multiple NFS requests against them).
Hmmm... This raises an interesting point. Yes, more operations should
happen, but the primary ref for the handle (taken by NFS4_OP_PUTFH) will
be once per compound, not once per op. So it would take multiple
compounds to advance to the MRU of L1. Not a problem for multiple reads
or writes, but if a file is opened and read/written once, and then left
alone, it won't advance to the MRU of L1.
Daniel
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Nfs-ganesha-devel mailing list
Nfs-ganesha-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
