Stripping disk to a unix host

Hi!

You can have two options:
1. Unix do the striping (Host side striping)
I think it is better option.
A few years ago emc suggest us if we want to
decide between striping on the host side (Windows) and Symm side, we should
use host striping.
The advantage of this solution:
- You can configure the striping on the host side, you dont need to make a config change on the symmetrix, you should make striped volumes with lvm commands.
(Easier to configure)
- You should have a good Volume manager software, which is fast.
- Most volume manager software enables to grow the volume

disadvantages:
- A bit more configuration required on the host side


2. use meta volumes on the DMX (Dmx do the striping)
In this case DMX do the striping. DMX makes a big volume (called meta)
from a standard volumes. Then the DMX stripes the data on the meta volume.

Advantages:
- You can use a big volume, you dont have to configure too much on the host side

Disadvantages:
- You have to make a config change on the DMX. (bin file upgrade)
- If you want to change the size (for example it grows), you
may extend the volume, if it can be do online, it depends on the
host Operating System, on the microcode version ... etc
- Not very good performance.

Regards:
Paul

Oh i forgot:
If you want stripe on the host side, you dont have to activate anything on the DMX.
Just you have to assign a few volumes to the host, the you can
make the striped volumes on the host side ...

Which host is it? (Operating system)
Which volume manager do you use? (Veritas, Disksuite, LVM ... etc)

Hello,

I could make metas without a config change, i think it depends on your microcode. Contact your CE for details.

LVM-meta have a bad taste when you are using TimeFinder or SRDF.
Because you have to use consistency groups to get a consistent state of your copy.
And this feature is optional.

When you want to expend Metas on Unix online you have to use TimeFinder BCVs. Without veritas volume manager you have to unmount the meta -> change partition table -> mount FS -> growfs
And so you have a downtime around 5 minutes + application stop / start.

There are a lot of factors to look at... Some hosts do not allow to map more then a specified number of luns .. how can you present a lot of TB to a single host with only 255 luns ?? ... Sometimes striping at the host side gives a lot more pain than advantages.. Sometimes striping at the host side does not allow you to migrate your data from a volume to another volume using the LVM itself. Sometimes you need to share a single FA between a lot of different hosts .. So you have to share the number of luns that you can map on the FAs between all the hosts.

Metavolumes made into the DMX aren't the only answer to all the issues I've mentioned above.. But they can be AN answer.

I think that there isn't a single answer .. as happens almost anytime you ask questions about performances, the only answer is "It depends"