Our plan is leveraging the existing NFSv4 operations, not extending it. However, we might consider some small extensions if they show a lot of benefit. For example, to make NFS protocol capable of appending a file in one compound, which we don't know how to achieve without changing the protocol.
On Fri, Apr 8, 2016 at 6:18 PM, Matt Benjamin <[email protected]> wrote: > Hi, > > I guess Bruce's point may be a question, is the NFSv4 COMPOUND operation > itself sufficient for your planned implementation? Is extending NFSv4 > COMPOUND/operations and semantics part of the project? > > Matt > > ----- Original Message ----- >> From: "Ming Chen" <[email protected]> >> To: "J. Bruce Fields" <[email protected]> >> Cc: [email protected], "Matt Benjamin" >> <[email protected]>, "Geoff Kuenning" >> <[email protected]>, "Dean Hildebrand" <[email protected]>, "Ashok Sankar >> Harihara Subramony" >> <[email protected]>, "Vishnu Vardhan Rajula" >> <[email protected]>, "Donald Porter" >> <[email protected]>, "Erez Zadok" <[email protected]>, >> [email protected], "Garima Gehlot" >> <[email protected]>, "Jasmit Saluja" <[email protected]> >> Sent: Friday, April 8, 2016 6:01:27 PM >> Subject: Re: [Nfs-ganesha-devel] NFS Transactional Compound Project >> >> On Fri, Apr 8, 2016 at 5:21 PM, J. Bruce Fields <[email protected]> wrote: >> > On Wed, Apr 06, 2016 at 10:37:51AM -0400, Ming Chen wrote: >> >> We have just created a github repo about our transactional NFS >> >> compounds project: https://github.com/sbu-fsl/txn-compound >> > >> > To me a "transaction" is an atomic operation that succeeds or fails as a >> > unit, which an NFSv4 compound isn't. >> > >> > I know that's not what you're aiming for, you're just looking for >> > performance improvements, but the name might confuse people (on a point >> > that people are already inclined to get confused about). >> > >> > --b. >> >> We actually meant "transaction", including ACID. We intend to build a >> NFS server that can process a compound as a transaction. The plan is >> to leverage BetrFS (http://oscar.cs.stonybrook.edu/projects/betrfs/) >> which is local FS built on top of a database engine. Don and Yang are >> involved here for that. Our lab has done related work before such as >> KVFS (http://www.fsl.cs.sunysb.edu/docs/fast-kvfs/kvfs.pdf) and Cosy >> (http://www.fsl.cs.sunysb.edu/docs/cosy-hotos/index.html) >> >> It is just that, at this early stage, we are focusing on the >> compounding part. Transaction is an option in the API. >> >> Best, >> Ming >> > > -- > Matt Benjamin > Red Hat, Inc. > 315 West Huron Street, Suite 140A > Ann Arbor, Michigan 48103 > > http://www.redhat.com/en/technologies/storage > > tel. 734-707-0660 > fax. 734-769-8938 > cel. 734-216-5309 ------------------------------------------------------------------------------ Find and fix application performance issues faster with Applications Manager Applications Manager provides deep performance insights into multiple tiers of your business applications. It resolves application problems quickly and reduces your MTTR. Get your free trial! http://pubads.g.doubleclick.net/ gampad/clk?id=1444514301&iu=/ca-pub-7940484522588532 _______________________________________________ Nfs-ganesha-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
