Yesterday was Cisco’s big day. The networking behemoth ended months of speculation about its move into the server market by unveiling its Unified Computing System, codenamed “California.” As I sat through yesterday’s video conference helmed by Cisco CEO John Chambers, I kept waiting for the nitty-gritty details of the technology the company is introducing. Instead, I heard more than 90 minutes of chatter among Cisco executives and leaders from partners like VMware, EMC, BMC, Accenture and Microsoft about how well all these companies were working together to execute Cisco’s vision. I was a little disappointed. The conference was much more conceptual than technical, and I think that’s the way Cisco wanted the day to go. Cisco PR folks blitzed editors and analysts with fact sheets and press releases during and after the event, and there is a lot of meat in there. It just takes a while to read through it all.
Cisco’s rivals, particularly its rivals in the networking market, are eager to offer their opinions on what Cisco is trying to do. Michael Morisy will take a deeper look into that with a story on SearchNetworking.com tomorrow. In the meantime, here’s some of the feedback I’ve received from them over the last 24 hours.
First up is Mike Banic, Juniper’s vice president of product marketing for Ethernet platforms. He notes that Cisco’s event was more conceptual than technical because Cisco may have been forced to unveil this project a little early.
There was a rumor that this event was happening a lot earlier than [Cisco] had planned because of the article in the Wall Street Journal on California. It looks like that reporter had a lot of good facts. So they moved this up because they didn’t expect anybody to get these details and they needed to slowly unfurl the story. They’d rather be telling this story when they have more details to share. i don’t think they’re really there. If they wante dto show us how they coulud simplify management [of the data center], they would demonstrate that. Otherwise, it’s just words.
Banic offered a contrast between Cisco’s approach and Juniper’s approach to simplifying management of data centers.
[Cisco is] not suggesting anything new here. Simplified management of compute resources and networks is something vendors in this market have been doing for awhile, like HP and IBM and newer entrants like VMware. For us, it’s very different from our strategy. We’ve focused on being network pure-play. The network is our strength. We’re not going to wander into knew worlds like servers. We’re focusing on connections. [Cisco] is going to perpetuate the multi-layer network model, whereas the Juniper vision is to have the whole data center network look like one switch. It will be multiple switches, but it will all look like one switch [in the management console]. We already have something like that with our virtual chassis switch. We can build a single logical switch for the data center. That’s what Project Stratus is.
Banic also noted that Cisco’s entry into the server market will further drive a wedge between it and traditional server manufacturers like IBM and HP, which Cisco has partnered with in the past. “Those vendors are in a better position and have more expertise and history with servers than we as a networking vendor. And we’re in a better position to work with those partners, like IBM.”
Brocade has been positioning itself to become a player in the data center networking market, particularly through its acquisition of Foundry Networks. Elizabeth Walther, Brocade’s senior public relations manager, offered me the following observations:
- Cisco’s approach to Unified Computing is not revolutionary. Many companies with extensive experience in solving complex data center issues are already working on solutions
- Cisco’s approach is likely to be very capital intensive up front, which will be a major obstacle in light of today’s global economy.
- The challenge at hand — the evolution of the data center to a dynamic, fully virtualized state — is extremely complex and should leverage open architectures and industry standards.
Her second point, that it might be expensive, is a valid point. But I think companies that are looking to transform their data centers in this way know that they will have to lay down money to do it. Also, Cisco isn’t expecting broad adoption of this technology until four or five years out. By then (we hope) the economy should be rebounding.
Brocade is arguing that, despite Cisco’s talk about using open industry standards in Unified Computing, the initiative will still involve too much proprietary technology. Brocade offered this official statement, which expands on that point:
A dynamic and virtualized data center holds the promise of many compelling benefits for end-users including increased server utilization, decrease in power footprint and more efficient operations in general. However, achieving this goal is a complex challenge that can be best tackled by a broad ecosystem of industry partners and not based on a proprietary, singular architecture of one company.
In contrast, Brocade is already helping customers address these challenges by integrating our networking solutions with a range of mature computing, management and storage technologies from some of the strongest companies in the world. These partnerships are leveraging open interfaces/standards, co-developed technology, and products that are available today, which will lower costs and maximize return on investment for customers.
Blade Network Technologies
Blade CEO Vikram Mehta echoed Brocade’s position on Cisco’s ideas of industry standards will lock customers into a “proprietary world” while locking out vendors like HP and IBM that are “trusted open systems suppliers.” He said the standards in Unified Computing are tantamount too “standards with a C” as in Cisco.
Methta trashed Cisco’s announcement in his own blog with 10 reasons why Cisco’s Unified Computing won’t fly. Here are three of them.
- Unified Computing means standards with a “C.” According to Cisco, converged data and storage networking requires Cisco’s Data Center Ethernet (DCE), thus eliminating freedom of choice with a sole-source Cisco-only server and network. This puts at risk integration and interoperability with vast existing installations. The rest of the industry is working on an open approached called Converged Enhanced Ethernet (CEE) using IEEE’s Data Center Bridging (DCB) standards.
- It’s more about packaging than true innovation. For example, Cisco’s fabric extenders carry the same cost structure as switches, as they utilize similar switching silicon, physical interface components, and management processors. When compared to traditional switches – sharing management via “stacking” – the fabric extenders are another example of packaging, not innovation. The more costly data center infrastructure components – CPUs, RAM, and networking silicon – remain unchanged, except Cisco’s prices are higher and – surprise, surprise – more Cisco gear is needed to control them.
- Follow the money – into Cisco’s bank account. Cisco’s “California” server approach requires Cisco’s Nexus 5000 switches that start at $17K for a bare-bones Layer 2 switch and significant premiums for adding Layer 3 and FCoE functionality, so the total cost of ownership will be similar to the cost of living in California.
I should note that Cisco said Unfied Computing does not require investment in Nexus switches. Cisco executives told me the uplinks from a Unfied Computing System can plug into any vendors switches. Of course, customers will get the best performance out of the system by plugging it into Nexus switches, which offer the “unified fabric” technology that Cisco is promoting with Unified Computing.
As I receive more input from Cisco rivals, I will post updates here.