This is the latest “keep you up all night” challenge that I am grappling with. VMware has some phenomenal tools for building enterprise computing utility foundations for virtual datacenters, and at VCE we provide the best platform for seamlessly integrating with the VMware model and providing dynamic capacity on demand and fully orchestrated resource management, but when you start to consider “N” scale and a self-service business model in the commercial space, things get hairy.

EDIT: worth noting that the maximums are hypothetical. reality will be far lower since there are various reasons (SCSI LUN contention, maintenance windows, host CPU utilization, LUN limits, etc) to not come near those hypothetical maximums.

EDIT 2: also important to note two things re: SRM – there are no linkages between vCD and SRM (so potentially no way to automate protection easily or carry over vCD properties in site recovery) and SRM had a limit of 1000 VMs

I am thinking that this design may be heading down the path of 3rd party orchestration and glue code/customization.

EDIT 3: Another limitation to work through is that UIM currently is limited to a 8 UCS chassis.  The Cisco UCSM domain limit is 20 chassis from an “officially supported” view, but Vblocks are orchestrated via UIM so it makes more sense to adhere to UIM scale limits and orchestrate multiple UIM instances.  Orchestrating UCSM directly would provide more scale from a chassis management standpoint, but would lead storage unconvered and would require higher level orchestration anyhow since 20 chassis would be unlikely to completely satisfy the entire lifespan of a carrier IaaS.