Announcement

Collapse
No announcement yet.

pNFS Block Server Support Is Coming To Linux 4.0 (3.20)

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • pNFS Block Server Support Is Coming To Linux 4.0 (3.20)

    Phoronix: pNFS Block Server Support Is Coming To Linux 4.0 (3.20)

    The nfsd changes for the Linux 4.0/3.20 kernel provide pNFS block server support...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Maybe XFS will get a bit of a boost development-wise now when RHEL 7 uses it as default.

    Comment


    • #3
      This means pNFS will be able to be a replacement of NBD??

      I ask for choosing the best in the future.
      Thanks.

      Comment


      • #4
        Originally posted by q2dg View Post
        I ask for choosing the best in the future.
        Thanks.
        Not really NBD is more like iSCSI. Only one user can access the disk.... I have used NBD in the past for mounting large storage on a small machine IE a router or vintage computer (Sparcstation LX in my case).

        NFS and hence pNFS are network file systems ... and can be accessed by multiple users at the same time. Its a very cool development though.

        Comment


        • #5
          Originally posted by cb88 View Post
          Not really NBD is more like iSCSI. Only one user can access the disk.... I have used NBD in the past for mounting large storage on a small machine IE a router or vintage computer (Sparcstation LX in my case).

          NFS and hence pNFS are network file systems ... and can be accessed by multiple users at the same time. Its a very cool development though.

          On the flipside, won't this limit you to clients that all understand the same filesystem? I've got a FreeBSD/Linux mix here, with some samba serving of the same files to windows users ... and I have a feeling I'm not the primary target for pNFS.

          Comment


          • #6
            It's comparable with ceph?

            Comment


            • #7
              Originally posted by tessio View Post
              It's comparable with ceph?
              Or Venti from plan9, or iSCSI - it's a block storage that requires the client OS to do something intelligent with the blocks (like, say, putting a file system on them). Which is perfectly nice and useful, though it's almost required that your clients all run the same OS.

              Of course, you could set up a file server that served the "difficult" clients from files stored in pNFS/ceph/whatever - blocks, I'm just not sure if it would be worth the extra management compared to just serving files directly from a RAID. It looks like Ceph could get you some extra redundancy in a way that's harder to do with fileservers, so that might be useful in certain cases?

              Comment


              • #8
                I'm confused that this can be a feature. Once you start letting clients access the datastore as a block level device, aren't you basically throwing all the features of NFS out the window and getting the equivalent of iSCSI?

                Comment


                • #9
                  Originally posted by pgoetz View Post
                  I'm confused that this can be a feature. Once you start letting clients access the datastore as a block level device, aren't you basically throwing all the features of NFS out the window and getting the equivalent of iSCSI?
                  Actually, never mind; I was confused. The whole point of pNFS is [to provide client's with metabacked block level access to storage. The actual question is how did linux have any pNFS support prior to this?

                  Comment

                  Working...
                  X