I have been presented with the need to create a virtual machine with a very large storage space. Right now its 4TB, but one day, it could be more. I am fishing for all the ways possible to make this happen. Some might be better performing, some may be more fault tolerant, but the best will be a high performing fault tolerant method.
I have one way working and testing right now, but I have concerns for fault tolerance.
My working Scenario:
We have a compellent SAN (Thin provisioning, Teired Storage, Fast Track, Virtual Volumes, etc) , I created 2 Luns with 1995 GB. I created 1 1.9G .vmdk on each lun and mapped to a virtual machine. Once in the windows 2008 vm, I took those 2 drives and did a RAID 0 stripe. Performance was fantastic at 550+MBytes/sec. Thats darn near the capacity of the 4G fiber. Drawback, there is no way to lash the LUNs together in such a way as to guarantee the stripe will stay intact on an outage.
I know I can make multiple LUNs and use vmware to create an extent/span, but im still limited to the .vmdk at 2TB-512M limit. So, I can still create 2 .vmdk files, but how resilient is this, especially when the SAN doesnt know they are merged at the ESX host level?
I need some feedback on how some of you folks have resolved this issue IN PRODUCTION and how fault tolerant it can get. I know some of you may use another brand of SAN other than compellent, but thats fine, just tell me what you use and what the feature you use on the SAN to make sure separate LUNs are safe in an outage.
PS. I should add that I am aware I can use a really large LUN with iSCSI initiator software inside the vm. I am trying to think outside the box for a bit. iSCSI may not perform as well as being able to use the 4G fiber option.
With a supporting HBA and supporting hardware platform you may be able to use passthrough for direct VM OS to SAN.