Create one storage group for every host that you have attached to the SAN.  If your SAN isn't growing, you'll never need to add more storage groups.

5-7 of the below document explains the planning in detail.

http://www.hecomputing.org/files/clariion/CS-Series_Config_Guide.pdf

on that same note, you want to create a single storage group for each cluster you have.  let say you have a 5 node ESX cluster, you would not create 5 separate storage group.  you would create one and place all hosts in that group.  That also helps alot because in clustered environments, the HLU# needs to be the same across all nodes in the cluster and that is easily set when you add all hosts for a specific cluster into the same storage group.

@kenn2347 is correct about clusters for most environments, but there are some exceptions.  All hosts in the same storage group will see ALL LUNs in that storage group which is great for clusters since you need all nodes to have access to the same LUNs, and you need them all to have the same Host LUN ID.

However, if you are booting-from-SAN, you will need to create a separate storage group for each node so that each node see's only it's own boot LUN.  Then the clustered LUNs can be added and you can assign the same HLU to each LUN in each storage group as you add the LUNs.  It's more work than using a single storage group for the entire cluster, but it's necessary in a boot-from-SAN environment.

As each node runs on its own OS. We do not share the boot luns between the nodes in a cluster. Say if we have 2 nodes , we create 2 boot luns and give it to each node individually ? Did I get it right?

Ok guys, offtopic: please don't call a storage array a SAN, because the two are completely different things. SAN is the network that connects storage to hosts; a SAN is the Storage Area NETWORK !

what took you so long ? hahah

In EMC-speak, its going to be SAN and NAS, because we think in terms of environments.

In particular, the new EMC documents for Unified, which is the merging of the CLARiiON and Celerra lines will be using SAN and NAS, because its cooler than block and file.

But, you're a Hitachi guy now, so you shouldn't be upset.

Whoahahaha, I'm not an HDS guy at all...... they let us down recently and we're very upset because they made a promise they couldn't keep. No way I'm HDS... I'm EMC !!!!!

There is at least one more exception ,, consider a Oracle RAC cluster with shared diskgroups , but individual local file systems for database binaries(orabin, oraadmin). Here only Disk Groups LUNS  can share the same storage group