One year ago Cisco was in turmoil. Wall Street was pouting, competitors were crowing and customers were scratching their heads. If you believed some reports, you might think the good ship Cisco was steering into the iceberg that Nortel struck a decade ago.
Last week Chambers huddled with eight tech journalists via TelePresence to update us on the progress of Cisco’s new course. We hit him with questions for about 90 minutes. Chambers is a polished executive whose presentation plays well with CIOs and Wall Street. My readers are network architects and engineers, yet his comments on Cisco’s overall strategy touched on some very relevant issues for the guys who push packets. Here is what he had to say.
No fear of merchant silicon, John Chambers still considers ASICs an edge
Industry buzz might be centered on software-defined networking over commodity hardware, but Chambers doesn't agree. In fact, he says hardware and custom silicon will continue to differentiate the company, and that ASICs (application-specific integrated circuits) will remain Cisco’s jam.
Chambers admitted that merchant silicon worries him now more than it did five or six years ago. Arista Networks, for example, has built its whole business on merchant silicon, devoting resources to software. However, Chambers insists Cisco won't go that route.
“The combination of software, hardware, services and ASICs -- and ASICs play a big part -- have to play together,” Chambers said. “That will be difficult for others to emulate. You will see us become more of a software company, but we think software with an ASICs and hardware combination ... we will cover our bets in both areas.”
So while Cisco continues presenting an ASIC-centric approach to the market, it will maintain skunkworks projects internally that allow it to move with the market. The low-latency Nexus 3000 series switches are a recent example of Cisco’s willingness to go with merchant silicon when the market dictates.
Is Cisco really on board with OpenFlow, software-defined networking and next-generation network virtualization?
Like many industry heavyweights, Cisco is a member of the Open Networking Foundation, the group that is guiding the development of OpenFlow-based, software-defined networking (SDN). But Cisco execs have said that OpenFlow isn’t quite ready for production and that SDN is much bigger than OpenFlow.
One can argue about OpenFlow’s readiness, but software-defined networking is catching on. Start-up Nicira Networks emerged from stealth mode a couple of weeks ago and announced its SDN-based network virtualization technology is in production with eBay, AT&T and Fidelity.
Where does Chambers stand on SDN, OpenFlow and network virtualization?
“We absolutely view it as a viable option for the future, either as a total architecture or segments of it,” he said. “We probably spend a couple hours a week focused on our strategy in this market through a combination of internal development, partnering and acquisition. If we do our job right you’ll see us move on multiple fronts here. And at the right time, when it is a benefit to our customers, we will outline our strategy for them.”
Here again, though, Cisco’s custom silicon will be key. Cisco will tie its SDN and network virtualization strategy to its ASICs, while maintaining a commitment to interoperability through open standards, he said.
In the meantime, though, companies like Google and Yahoo have made it clear that they need solutions like OpenFlow-based SDN today, and they are building their own SDN solutions with networking gear from little-known vendors that have embraced OpenFlow, like Taiwan’s Pronto Systems.
Chambers wouldn’t comment on specific customers among the web giants, but he admitted that there are different views on the state of SDN solutions today.
“We are focused very heavily on massively-scaled data centers,” he said. “We will not have a technology religious view. We will make our bets on both ways [OpenFlow and other network virtualization technologies]. We bet more heavily on which one we think might win and in which sequence. But I think it’s fair to say that we’re doing better in those environments than we were just a year ago.”
John Chambers on the QFabric buzz
If 2011 was the year Cisco battled hard against HP Networking, 2012 will be the year that Cisco goes to war with Juniper Networks over QFabric. During his talk, Chambers took some indirect shots at Juniper for how long the company took to roll out QFabric (Juniper made the solution available late last year after first making a vague announcement of “Project Stratus” three years ago).
“We’ve been very effective in making commitments -- for the most part but not always – on delivering on timeframes we set for our products and our directions,” Chambers said.
But how will Cisco counter QFabric now that it’s here?
If Chambers’ comments are any indicator, Cisco will play up its larger data center architecture, including its servers (the Unified Computing System), its technology partnerships (vBlocks via VCE), and services.
Read more Fast Packet bloggers
HP's OpenFlow is a move away from the “tyranny of CLI”
Are networking vendors doing enough to prevent software bugs?
With edge software overlays, is network fabric just for raw bandwidth?
“The architecture that [Juniper] is working on is very interesting,” Chambers said. “This is an industry… where you tie what you do in the data center to what you do in the cloud, to what you do in the enterprise environment, to what you do in the service provider environment. We think these pieces will come together. I believe the biggest transition of all will be how all these pieces work together as opposed to a single product.”
And what about those web giants that are turning to commodity network hardware?
“If you are in a market where you add dramatic value through your product, you have very little interest in playing in a commodity markets,” Chambers said. “You will not see us put a lot of resources into something that will be commodity, that we can’t differentiate, and that doesn’t tie together with smart networks. On the other hand, we have a huge interest in bringing higher functionality to any device over any combination of networks from which these massively scaled data centers will occur. Designers of data centers really care about how they get their capabilities out in a way that gives them a competitive advantage. If we can do that through an architectural play, then we’ll do fine.”
John Chambers: Mistakes were made, so we reinvented ourselves
As for all the bad press a year ago, Chambers says it wasn't a result of renewed competition from the likes of HP and Juniper, but rather a drop off in government technology spending and a problem with Cisco hitting too many market adjacencies at once. The organizational changes that Cisco has made in the past year should have been made earlier, Chambers admits.
“I should have had a COO a year earlier,” he said. “The effectiveness of people is how you lead your own organization, how you do collaboration across groups, how you voice disagreement in a way that gets the desired behavior. If you watch [COO] Gary Moore, he is world-class on getting messages across. You’ve got to get people and customers and press around you that can challenge you, but back it up with logic behind the challenge in a way that gets the desired action.”
Cisco also misplayed its hand in certain markets, particularly with its network security strategy. Customers were grousing over Cisco’s lack of focus for a couple years, and Cisco is still in the early phases of turning things around, Chambers admitted.
“We did not have a strategy ... in the one area that is subject to an architecture [more] than anything else: security,” he said. “If you watch what [Cisco Senior VP of Cisco Security Group] Chris Young does for us in security, you’ll look back in a few years and say it truly is an architecture approach.”
Other Cisco strategies have met with some criticism, particularly
Borderless Networks, which some engineers have panned as marketing hype. Chambers disagrees,
saying that Cisco may need to do a better job of communicating its architectures to the market, but
he called Borderless Networks a tremendous success. Whether that's true is hard to gauge. Let’s see
how long it takes for Cisco to roll out a replacement brand.
Let us know what you think about the story; email: Shamus McGillicuddy, News Director