Hi, Is ZFS comparable to PVFS2? Could it also be used as an distributed filesystem at the moment or are there any plans for this in the future? Thanks and best regards, Ivan This message posted from opensolaris.org
Ivan wrote:> Hi, > > Is ZFS comparable to PVFS2? Could it also be used as an distributed filesystem at the moment or are there any plans for this in the future?I don''t know anything at all about PVFS2, so I can''t comment on that point. As far as ZFS being used as a distributed file system, it cannot be used as such today, but it is something we sould like to develop. Do you have a specific use case in mind for a distributed file system? -- --Ed -------------- next part -------------- A non-text attachment was scrubbed... Name: ed.gould.vcf Type: text/x-vcard Size: 282 bytes Desc: not available URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20070108/a0ca4133/attachment.vcf>
Hi Ed, pNFS (Parallel NFS) could benefit by using a ''distributed filesystem version'' of ZFS. By using pNFS files could be striped along different NFS servers. Lisa Week (lisa.week at sun.com) told me that they would like to use ZFS in future pNFS Servers in Solaris. Thanks and best regards, Ivan This message posted from opensolaris.org
They''re really not comparable at all. ZFS is a local on-disk file system. PVFS2 is a parallel file system built on top of a file system local to each of the nodes in a cluster with a lot of features for high-performance I/O. This message posted from opensolaris.org
At the september LISA meeting Jeff B. did suggest that they planned to - eventually - add the distributed aspect to ZFS, and when he''s talking about the filesystem as a ''pool of blocks'' it certainly seems like there''s no reason (beyond some minor implementation issues :) why those blocks could not reside on different hosts. I have not looked at the ZFS implementation closely enough to know how it deals with protection of the inodes and directory structure (ie. if I lose a disk or host can I reconstruct the inodes and directory tree) so this may introduce some complexity if you want your filesystem to survive one- or multiple failures. I would''ve assumed this has been thought of, but after discovering that I can''t add a disk to my raidz2 pool (but it''s just a pool of blocks!) I''m slightly less confident of this .. This message posted from opensolaris.org