Cisco to reveal next 'Data Center 3.0' push
Seeding the inter-clouds
Server wannabe Cisco Systems is on a five step program to take over the data center with its Data Center 3.0 initiative, and on June 30, the company is getting set to divulge what the next step will be – and perhaps announce a product or two.
The Data Center 3.0 roadmap is something Cisco's top brass have been talking up for years, first as "data center networking," then "unified fabric" (with the Nexus switches) and then "unified computing," with their "California" two-socket B Series blade servers and their integrated switching coming to market in March 2009. These were the first machines to launch with Intel's quad-core Xeon 5500s, putting the upstart server maker at the head of the PR and marketing line.
A year ago, Cisco fleshed out the Unified Computing System with the C Series rack-based servers (also using Xeon 5500s) and earlier this year added with four-socket blade and rack machines using Intel's eight-core Xeon 7500s while also qualifying the six-core Xeon 5600s on its existing blades and racks.
That's the first three phases of Cisco's attack on the data center. Phase four was to take the UCS systems and the combination of Nexus switches and C Series rack servers (for those who want racks instead of blades) and build private cloudy infrastructure. This was arguably already launched as the Vblock products that are being sold by the Acadia partnership between Cisco, EMC, and VMware. The Vblocks, which were announced last November, come in three different sizes and incorporate Cisco servers, EMC storage, and VMware virtualization tools that support as little as 300 to as many as 6,000 virtual machines.
That leaves phase five still undone, which Cisco calls "inter-cloud" and by which it means allowing for cooperation between private, internal clouds and public, external clouds. And presumably cooperation among public clouds too. The idea is to have workloads be portable (at least as long as they run on the x64 instruction set).
A lot of the work for the inter-clouds has been done in terms of the UCS iron supporting different virtual machine architectures. VMware's ESX Server hypervisor is by far the dominant one on x64 iron and is the preferred hypervisor for California boxes, other hypervisors are supported on the machines, as you can see here, including Citrix Systems' XenServer 5.5, Microsoft's Hyper-V 2008 R2, Oracle VM 2.1.5 and 2.2, and Red Hat Enterprise Virtualization (the commercial-grade KVM). Various provisioning tools from BMC, CA, IBM, Microsoft, and Symantec are also already supported.
But the inter-clouds will require more than supporting hypervisors and management tools. For applications to truly be portable on clouds, you need to be able to convert running workloads from one VM hypervisor (including its file system format and management tool hooks within that VM) to another. And what the heck, let's also add in the requirement that you also need to be able to live migrate a running VM as part of the inter-cloud process.
Cisco is not divulging what it will be talking about on June 30, but El Reg will be listening in to hear all about the inter-clouds and also to get an update on how UCS sales are going.
Hopefully there will be some new server iron too. It would be very interesting to see Cisco adopt some Opteron processors, particularly for customers who have thermal issues, but that seems like a pretty remote possibility. Cisco is a very tight partner with Intel and has never even hinted that it was looking at Opterons.
Cisco could field an eight-socket C Series rack using the Xeon 7500s, but that also seems unlikely unless customers have been crabbing that 32 cores are not enough to support database workloads. The existing four-socket B Series blade and C Series rack machines are probably enough for lots of customers. Cisco tends to shoot for the belly of the market and not work the fringes of the bell curve. ®