Cisco throws California virt-server gauntlet
Much talk, action pending
The gauntlet has been thrown. Cisco Systems launched its "California" Unified Computing System today in a 14-city IPTV extravaganza, with the key partners it has tapped to provide components to the system showing up to talk - and talk a lot - about what this all means. Pity that all this talk was a little light on the details about exactly what a California system is and what it will cost.
It is now pretty obvious why Cisco chose to do its announcement today, two weeks ahead of the expected launch of the quad-core "Nehalem EP" Xeon processors for two-socket servers. It wants CIOs to give pause before committing to buying new rack or blade servers from the usual suspects: Hewlett-Packard, Dell, IBM, Sun Microsystems, Fujitsu, and a handful of others who will be launching Nehalem boxes this spring.
It is not clear yet when the California system - and Cisco's top brass is emphatic that it is a system, not a point product like a blade server - will be available or what specific components it includes. At least not based on the launch today. Cisco is going to make us work a lot harder for the actual feeds and speeds.
First and foremost, and contrary to some rumors last week, the California system does indeed contain blade servers, and these are blade servers that have been designed by Cisco using Intel's forthcoming Nehalem Xeons. The blades may even be manufactured by Intel, for all we know right now. But John Chambers, Cisco's chief executive officer, was emphatic (well, in his calm way of being emphatic) that Cisco is not interested in being perceived as jumping into the blade server market and competing with the likes of HP, IBM, Dell, Sun, and Fujitsu.
"We focus not on competition, but on where the market is going," Chambers explained, saying that the "biggest market play" that Cisco could identify was to create a system that encapsulates compute, storage, and network infrastructure and the virtualized instances of that technology.
He said that what Cisco does is lead and that "leadership is about how you catch market transitions and inflection points." Chambers admitted that the best technologies do not always win in the marketplace but the best strategy nonetheless was to create technology and let "your customers help steer you through it."
If you wanted to translate that into plain American, what Chambers might have said is that this consolidation of networking, serving, and storing was inevitable, and Cisco really had no choice but to compete with any server market for the IT dollars. And the server makers will have no choice but to push back too. Regardless of the sweet talk coming out of Cisco.
I liked this one from Chambers during the California system launch: "We are not taking a dollar of revenue and splitting it up differently, but taking a dollar of revenue and making it into two or three." This is the Way of the Internet, apparently, where you make pies bigger and you don't have to compete. In 2009, this kind of talk is just silly. IT budget growth is stifled and will likely go negative, and IT departments are going to be told to do more with less. Again. And Cisco is absolutely going to try to compete for the high-end, virtualized projects that IBM and HP in particular were counting on to make their numbers.
In any event, three years ago, just before Cisco made an investment in Nuova Systems, which the company spun in nearly a year ago to deliver its Nexus 5000 Fibre Channel over Ethernet switches, the company knew it was going to have to bring computing into its Unified Fabric switching and routing technology for networks and storage, and it knew that virtualization was going to have to be a big part of it. Hence the two investments in server virtualization juggernaut VMware, which is itself owned by disk array maker EMC, also a key California system storage partner (along with NetApp).
Chambers knows that Wall Street has been wondering what Cisco is doing jumping into the cut-throat x64 server space, and he explained the situation piecemeal as he talked. "We don't enter markets where we can't sustain competitive advantage and margins in," he said. "We enter markets when we see inflection points."
The idea is that by combining servers, storage, networks, virtualization, and system management into a single, unified system, Cisco and its partners can chase some big deals where virtualization for Windows and Linux workloads on x64 iron is expensive to do even if the individual pieceparts are fairly inexpensive. The California system, according to Cisco, will support thousands of virtual machines and will require fewer servers, switches, adapters, and cables than current machines, and have more main memory and fewer I/O bottlenecks, thereby improving bang for the buck.
Two key parts of the California system come from partners VMware and BMC Software, the latter of which makes a living selling tools to help system administrators cope with myriad kinds of servers, operating systems, middleware, and applications. Cisco has inked an OEM agreement with VMware, as it turns out, and will be integrating its hypervisor and unspecified virtualization management tools into the California system.
VMware has also worked with Cisco to create the Nexus 1000V virtual switch, which as the name suggests virtualizes the links between VMs running on servers and physical Cisco switches such that as VMs teleport around the network, their network links are not broken. The added bonus is that this big switch is an instance of the Cisco IOS operating system that is actually running inside an ESX Server VM on an X64 server - presumably on one or more of the blades in the California setup.
Cisco has, as rumored, tapped BMC Software as an OEM partner for managing everything above the iron, and has OEMed BMC's BladeLogic software for this task. The Unified Computing System Manager software that Cisco has created to manage the hardware and that is embedded in the 10 Gigabit Ethernet fabric of the system feeds all kinds of data through XML APIs up to the BladeLogic software, which is used to provision, patch, and manage the operating systems and software stacks running inside VMs.
Microsoft is, of course, a partner on the California system, since you can't ignore Windows in the data center, and presumably, Hyper-V will be supported alongside ESX Server on the hypervisors. (No one at the Cisco launch answered that and many other questions seeking details). On the Linux front, Red Hat and Novell are partners and their Linuxes will be deployed as part of the system, presumably on ESX Server at first and then maybe atop Hyper-V since Red Hat and Novell both have interoperability efforts in the works to make sure their respective Enterprise Linux and SUSE Linux Enterprise Server distros can run atop Hyper-V.
There's obviously a services play here too, since this is an integrated system, and we learned today that Cisco and Accenture have formed a joint business group to help customers suss out where they can deploy California systems and help them do it. CSC has also been tapped as a Cisco system integration partner, and Wipro and Tata have also been asked to partner to help with application and business process re-engineering as it relates to customers moving to California systems.
It seems unlikely that other server makers are going to be invited to the party to plug their blade infrastructure into the California line, and when Chambers was asked about this, he deflected the question and another executive, Rob Lloyd, who was just named executive vice president of Cisco's worldwide operations (effective at the end of April) a few weeks ago, basically didn't address that issue and said repeatedly this is going to be sold as a system.
That sure doesn't sound like a BladeCenter or BladeSystem will slide in there. The integration of the management tools, the memory extension technologies Cisco has created for its Nehalem blades, and other unique features will make it difficult for Cisco to truly open up the system if it wants to keep integration and support costs down.
When asked about how big of a market opportunity this might mean for Cisco, Chambers hedged a little. "I always like to let products run for a year before we talk about what is possible," he explained, adding that the California system would now allow Cisco to chase about 25 per cent of spending in data centers. That's a lot bigger piece of the pie than Cisco gets now, and he went on to say that the California system could end up being a $10bn opportunity for Cisco some years from now.
The one thing that Cisco is clear on is who is signing off on these deals: the CIO. Cisco and its partners are going right to the top to push the California systems, right over the heads of server, storage, and network managers who want to protect their own fiefdoms. "It will be an executive engagement," Chambers said.
And you can bet that Cisco is going to be pushing the fact that it can support virtualized x64 workloads with about half the infrastructure as competitors are cobbling together and reduce both capital and operating expenses by at least 35 per cent. Those are the kinds of numbers CEOs and CIOs like to hear even before they get the feeds and speeds. Which is why Cisco talked about them in that order. As soon as the machines are shipping, Cisco will move on to talk about how the machines are intended to create private clouds, and then after that happens, Cisco will talk up something it is calling "inter-clouds."
Cisco says that it has ten beta customers using the California systems and has deployed it in production in its own IT shop; Pat Gelsinger, who manages Intel's Digital Enterprise Group and the partnership with Cisco, says he just told Intel's IT group to roll it out, too. Cisco says that it will have early customer trials "next quarter" and did not provide a date when the system would be generally available.
As we learn more, we'll tell more. ®