Up until now Cisco along with other vendors have been hugely driving up memory capacity while maintaining low CPU counts, Cisco’s Extended memory technology certainly gave them the edge in this space, however with greater DDR3 DIMM sizes ( 16GB now supported across all half width UCS M2 models) this advantage is not the differentiator it once was, as other vendors and non EMT Cisco UCS blades have now caught up or in some cases surpassed their EMT counterparts. There is still a case to be made about higher clock rates for Nehalem/Westmere EP processors over Nehalem EX and EMT memory real estate allowing greater capacities to be reached with smaller DIMM sizes perhaps.
One of the major reasons that platforms are increasingly featuring high memory / low CPU-count configurations is that it played right into the sweet spot of VMware’s VSphere 4 licensing model, in which licenses are based on number of sockets and number of cores on those sockets. However the licensing model of VSphere 5 now puts additional licensing requirements on larger memory capacities. More specifically vSphere 5 licenses are based on “vRAM:” the total amount of memory dedicated to each virtual machine,
For example A dual socket server with 256GB of RAM would under the new licensing model roughly double the cost of your VMware licensing.
VMware’s customers were so outraged at this licensing change that this week VMware announced a revised model (as shown below) which does significantly lessen the blow by doubling the amount of “Pre Tax” memory for Enterprise and Enterprise + licensing models
Below is a comparison of the previously announced and the new vSphere 5 vRAM entitlements per vSphere edition:
while I do not think VMware intentionally targeted the UCS it has the potential to be a major factor in customers buying decisions.
I think while VMware is still clearly the market leader and the “Hypervisor of choice” for Cisco UCS and Enterprise servers in general for that matter, this new licensing model will I’m sure put additional focus on alternative Hypervisors primarily Microsofts Hyper V , RedHat KVM and Citrix XEN.
Interestingly enough while writing this my memory is now recalling that UCS 2.0 (Capitola) due September has added additional functionality around RedHat KVM (specifically VM-FEX support, which essentially replaces the hypervisor switch with hardware switching on the UCS Fabric interconnects). Did Cisco see this coming and have decided to switch hypervisor allegiances, who knows.
Of course, you’re missing the column in your table showing “Pre version 5” memory entitlement, which would be ∞