In a matter of hours, Cisco Systems (s CSCO), the San Jose, Calif.-based maker of switches and routers will announce its much awaited lineup of server products, furthering its theme of unified computing. The company has planned a splashy event tomorrow featuring Intel (s INTC) CEO Paul Otellini and VMWare (s VMW) CEO Paul Maritz.
We had first reported on Cisco’s server plans almost a year ago. It’s a logical step, given that the company launched Nexus data center switches and made a substantial investment in VMWare (s VMWR).The introduction of its servers — a move that is likely to put Cisco in conflict with partners such as IBM (s IBM), HP (s HPQ) and others — is part of an ongoing effort at Cisco to find new revenue opportunities that go beyond switches, routers and wireless devices. Cisco has identified many new areas ripe for revenue expansion, and the data center is one such area.
Doubling Down on Data Center
To put it simply, our world is getting more digital. It means more information is going online, thus placing new kinds of stress on our Internet infrastructure. From storage to servers to routers, everything is being scaled up to meet the needs of a post-Facebook world in which we create all sorts of digital pollution.
This rapid growth of digital information has resulted in the formation of megacomputers — Google (s GOOG), Salesforce (s CRM), Facebook, Yahoo (s YHOO), Amazon (s AMZN) and Microsoft (S MSFT) — which are essentially data centers connected to each other via massive fiber networks.
A similar trend has started to emerge inside corporations. Take the beleaguered financial sector as an example. The chaos unleashed by the subprime mortgage crisis has resulted in the formation of super banks such as JP Morgan Chase (s JPM) and Wells Fargo (s WFC). In order to get on sounder footing, they will (sooner or later) need to make their digital infrastructure intelligent and use their mammoth data repositories by analyzing the information and turning that into a key business advantage.
At the same time, the geographically and demographically diverse nature of global corporations is forcing companies to move their business processes online. Many will have no option but to turn to the web and web-based software to unite its far-flung operations because they would need to be operationally efficient. The current global economic crisis has taken away the luxury of operational inefficiencies, regardless of the size of the company.
Get Ready for Comm-puting.
As companies lean more heavily on digital data to support their operations and performance, they’ll need new data centers that can handle the needs of a real-time corporation. To build this infrastructure, corporations will have to switch away from the old enterprise class hardware to more service-provider quality equipment. This is the long-predicted union of computing and communications into essentially comm-puting.
The folks at Sun Microsystems (s JAVA) were right when they said that the network is the computer, though they haven’t been able to capitalize on their core beliefs. Cisco, with no legacy in the data center business, has had to start from scratch, and thus has put together a solid strategy for the new emerging class of “comm-puting” equipment.
An essential quality of this new category is that it treats compute and storage resources as part of a single network fabric. Typically in data centers you would have three distinct resources — storage, compute (servers) and networks. Storage resources would be networked using different networking protocols than networks that tied the servers together. Cisco is taking a different approach (much of which will be outlined tomorrow). A key centerpiece will be “Project California” as reported by Internet News, which also has some details on the new servers:
- The new machines will be focusing on networking and memory.
- Each blade will have two 2.93 GHz Intel Xeon x5570 (Nehalem) processors.
- These new processors have QuickPath Interconnect (QPI) which gives them a bandwidth throughput of 32 Gbps.
- 384 GB of DDR3 memory per blade. Typical blade servers have 128 GB memory.
- They will be 4u high and fit up to eight per chassis.
- Project California box will have a rack that can hold eight “blades” and seven will use server virtualization software from either VMware (VMW) or Microsoft (MSFT). The eighth blade will be a networking switch called Nexus. (via BusinessWeek)
No Me-Too Product
These technical details reveal a blade server that is essentially the “Incredibles” version of today’s blade servers. Last week, Cisco Systems’ CTO Padmasree Warrior told me that the company was not interested in building a me-too product but instead wanted to build an entirely new kind of a product.
“We want to unify all elements – network, compute, virtualization into unified computing.” (She explains Unified Computing in this video.) In addition, she pointed out that — in sharp contrast to server-centric data centers — the network would essentially become the orchestrator, though she didn’t quite elaborate on how this will be achieved.
Cisco’s move is well timed. The demand for servers inside the data centers is on an upswing — about 50 percent of the 8 million servers sold every year end up in data centers, according to a BusinessWeek report. Similarly more devices inside the data center means more power consumption, which shoots up the operational costs of the data centers. According to Intel, nearly 25 percent of the costs associated with running big data centers can be traced back to power consumption. As we reported earlier, much of the power is drawn by pizza box style servers.
“Server virtualization has lowered the capital expenditure, but the op-ex has gone up, and what we have now is islands of virtualization,” Warrior told me, pointing out that Cisco’s new “unified computing” efforts are going to be accompanied by new software management tools as well. “The pre-virtualization era was about hardware, and the post-virtualization era is about combining hardware and software.”