plex
Hi,
I like to see why vxvm introced the plex object and what are the plex advantages if you compare with lvm.
I mean in lvm when we create an volume ito an volume group,here is not an intermediate layer between physical disk(included in volume group) and the volume itself.On vxvm we have the subdisk->plex->volume.
Maybe using plex thr things are more controllable over lvm model.Or using subdisk we could construct un volume who uses better the I/O then using just lvm.
thanks so much.
Simple reason I would say is more logical arrangement & control is possible because of plex. If you are thinking of a simple volume than probably it would look comparable to any LVM however think about layered volumes, where 1 volume may have multiple subvolumes. (mirror-stripe, stripe-mirror, mirror-concat, concat-mirror), these type of complex volume configurations are logically arranged because of flexibility provided by plex.
G
Hi,
As Gaurav said, it's nessicary for complex situation and provide more possibility for the structure in volume level.
As my understanding, it brings more logical structure for the volume could have own RAID likely layers like mirror-stripe, stripe-mirror, mirror-concat, concat-mirror. And with plex, we could splite volume and form new DGs without moving data from physical layer. I'm not sure LVM could do such jobs. But as I know like NetApp WAFL is a VxVM/VxFS likely system which also provide the ability to move data into new logical group without physically moving data.
So plex do make confuse for the person get used to LVM, but it really wonderful to bring many new features. Hopefully you can agree with me after you find out.
Cheers.
I think the original main reason is that in Vxvm, each plex can be setup different so you can create a volume with one striped plex and one concat plex. You can also setup preferred plex policies for reading, so that if you have unequal plexes then you can choose faster plex for reading from and you can also set preferred plex to "site" so that when you mirror across sites, VxVM will read the local plex.
Later on in VxVM versions, layering was introduced and as has already been said, the plex object is required to show these more complex volumes.
With LVM, each plex is identical so they are called mirrors or copies of the volume, but each individual one is not given a name, since they are all the same.
When you break a mirror you can specify the disk, so no real need to know the plex name for this, but it does give you an alternate approach to remove by plex name, rather than disk name.
Mike
