Bridging datacenters with VMware vSphere and IBM SAN Volume Controller – Part 1

A few months ago I had the opportunity to be involved in an interesting project.  We wanted to demonstrate a use case for using a SAN Volume Controller cluster with VMware vSphere.  The catch was that the clusters (both SVC and VMware) would be installed in distance separated datacenters.  Luckily both products already supported a configuration like this, we just had to put them together and do the work to test it.

Since SVC code version 5.1 a “split I/O group” or “stretched cluster” has been a supported configuration.  If you’re not familiar with SVC then the concept of stretching or splitting it will probably be foreign.  So a little bit of background first….

The base building block of the SVC cluster is an I/O group, which in turn is formed by two SVC nodes.  A node is basically a modified 1U IBM System X server, however since availability and redundancy is an integral component of SVC, nodes are always installed in pairs.  Check out the following link to see the specs of the current SVC nodes.  http://www-03.ibm.com/systems/storage/software/virtualization/svc/specifications.html

Picture of SVC nodes:

In the smallest SVC cluster installation there is one I/O group.  The I/O group functions as a single unit where volumes are accessible from either node and there is redundancy.  Typically the I/O group would be installed in one datacenter.  However in a stretched configuration the nodes from the I/O group are installed in different datacenters.   This provides the same functionality as a “standard” installation including access to volumes from either node.

So what makes a stretched SVC cluster different?  Aside from the ability to install in two separate datacenters…the good news is not much!  I’ll cover stretched SVC in depth in my next post (Part 2).