One major selling point for UCS is its scalability. But how scalable is it really?
You have a double point of failure (the FIs) that manage all of the network and storage communication for hundreds of blades and potentially thousands of servers.
The supposedly “agile” UCS service profile requires at a minimum for the following to be interoperable:
1) FI firmware
2) Blade firmware
3) ESXi fnic/enic driver versions
4) Legacy SAN OS versions
5) Blade model interoperability with the FI firmware version
All of this is required to be compatible with one another just to maintain some level of stateless computing (the ability to swap out a blade and maintain OS consistency).
Of course in the real world, things aren’t so simple: older blade models such as the B200 M3’s are not compatible with the latest version of the FI firmware. You can’t upgrade the legacy SAN without upgrading the ESXi fnic/enic drivers and the blade firmware. You can’t upgrade the blade firmware without also upgrading the FI firmware versions.
See where I’m going with this? That is just for the hardware and hypervisor to be able to communicate with the legacy SAN – this doesn’t include upgrading ESXi, or vCenter, or SRM, or any of the other virtualization components that also must be interoperable.