Re: Improving IOPS

Gandalf Corvotempesta <gandalf.corvotempesta@xxxxxxxxx> · Sun, 6 Nov 2016 14:16:15 +0100



    Il 06/11/2016 13:28, David Gossage ha scritto:

    
          I see maybe you don't really means
            raidz1 here.   Raidz1 usually refers to "raid5" type vdevs
            with at least 3 disks otherwise why pay a penalty for
            tracking parity when you can have a mirrored pair.  So in
            your case you are changing it from one zpool like was laid
            out to multiple zpools with each one being 1 mirrored vdev
            pair of disks? 

          
    I'm really sorry. I've mixed RAID1 with RAIDZ1 that means RAID5.

    Sorry for that. Let's talk about 'standard raids': I mean RAID1
    
      
            So moving from a replicated to a distributed-replicated
              model?  or a striped-distributed-replicate?  what is the
              command or layout you would use to get to the model you
              are wanting to use?
          
        
    I'm trying to say that the current Lindsay solution could be better
    (AFAIK):

    Instead of using a single RAID10, where in case of a "mirror"
    failure you have to resync the whole node from the network (24TB in
    my example), 

    a RAID1 solution (with 6 RAID1) is better. In case you loose a
    mirror, you have to resync only that mirror from network because.

    
    Instead of having a plain replicated setup:

    
    server1:brick1, server2:brick1, server3:brick1

    
    you'll have:

    
    server1:brick1, server2:brick1, server3:brick1

    server1:brick2, server2:brick2, server3:brick2

    server1:brick3, server2:brick3, server3:brick3

    
    In a distributed replicated setup.

    
    Each brick is a RAID1 mirror. The aggregation like a RAID-0 is made
    by gluster

  
_______________________________________________
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
http://www.gluster.org/mailman/listinfo/gluster-users