A/B Fabric’s are an important part of designing a 5 x 9 datacenter, but for some reason everyone wants to cross connect their storage switches (Be it iSCSI or even FC). Here’s my thoughts on this.
The largest and cheapest way to insulate yourself from downtime is design things in such a way that a single stupid action by an administrator has a rollback procedure that doesn’t involve a console cable. I’ve seen people do stupid things with STP, drop VLAN’s critical for storage traffic, or turn on debugs that crashed IOS. All of these things are harmless if you have a second path that is actively being used (why I like Active/Active Symmetric SAN controllers). As soon as you do something stupid, your vSphere alerts of path failure should light up like a Christmas tree. For added simplicity of figuring out which side of the fabric has crashed, you can create a simple tiny Datastore called “A” and one called “B” so in larger orgs your vSphere administrators should be able to very easily tell you what happened.