The footprint of the data center is starting to collapse in the same way that servers have collapsed into VM's on a grand scale. The question is, who has the best "datacenter in a box"? Well, I'm not going to say that Cisco does, but the release of Cisco's second generation UCS server blade, codenamed "Ventura", might turn some heads, and here's why.
If you own an existing blade chassis from HP or IBM (which I do), the biggest limitation you run into when doing server virtualization on a grand scale is memory. Current generation dual socket blade servers generally support up to 12 DIMM slots at full performance. What makes Cisco's new server blade interesting is that on a dual socket blade, Cisco has managed to stuff 48 DIMM's slots on the system board, and it's able to address all of those slots at full speed using a technique that maps 4 DIMM slots to 1 using Cisco's "Extended Memory Technology".
From a cost perspective, the ability to stuff a server blade full of cheaper 4GB DIMM's as opposed to 8GB has significant value. However, should you choose to stuff a Ventura blade with 8GB DIMM's, you're looking at the ability to scale up to 384GB of addressable RAM on a dual socket blade. If you generally allocate around 2GB per VM, as I do, that's ALOT of VM's that you'll be able to pack onto a server blade. While I wish I had a UCS chassis in the InformationWeek labs to test with, I suspect that CPU will become your bottleneck as you pack all of those workloads onto a given Ventura blade, not a bad problem to have.
Agile computing is clearly a fascinating and emerging space that will most certainly change the way we architect our datacenters as we transition into the next decade. Are you familiar with the blade level advancements that IBM, DELL/Egenera and HP are making, or do you have an opinion on Cisco UCS vs. HP vs. IBM vs. DELL/Egenera? I'd love to hear your take, share it with us here.