X
Tech

Conversation about UCS with Todd Brannon of Cisco

Cisco discusses some of the thought behind the UCS family of systems. It's my view that making a mainframe out of X86 system components is one of Cisco's goals.
Written by Dan Kusnetzky, Contributor

When hardware suppliers publicize another research firm's findings, I seldom take notice. After all, each research firm has its own definitions of market segments; data gathering and analysis methodology; and so, the findings are typically different and very hard to normalize. So, when Cisco's PR firm sent me a note telling me that Cisco's Unified Computing System (UCS) now has 10,000 customers and then went on the claim that Cisco had captured the number three position in the X86 blade server market.

I immediately replied to the PR person with a few questions about the market share claim. Was Cisco claiming the number three position for shipments or revenues? Was that claim for a specific geographic region or worldwide? What was the time frame for that finding (a specific quarter or annual)? What research firm's reports included that finding?

I was informed that it was worldwide shipments for a specific quarter and the findings were part of IDC's Server Tracker research product. When I peppered the PR person with more questions about that finding, a meeting was scheduled for me with Todd Brannon, Cisco's UCS Product Data Center and Virtualization manager. Although our conversation started with a review of IDC's findings, what followed was far more important.

Todd and I got in a detailed discussion of the philosphy behind UCS. Here are a few points from that interesting conversation:

  • Distributed systems are more complex and potentially harder to manage than mainframes
  • If set up properly, distributed systems can approach the performance of certain mainframes, but this can be difficult for some organizations
  • As X86 microprocessor; Windows and Linux operating system; virtual machine software; networking technology and storage technology evolved, companies were increasingly deploying complex distributed systems and complaining about the complexity and cost of operations and management
  • Cisco decided to develop a set of integrated systems that pulled all of the components into a single cabinet and management software that would offer mainframe-like management capabilities to address both the complexity and operational issues
  • Because these systems were easier to design, install and manage, they were seeing increased rates of adoption.

Although Todd didn't directly address performance, management or cost comparisons between the best of what Cisco had to offer and the best of suppliers, such as IBM had to offer, he seemed to believe that if such a comparison was made, his products would show well.

I have to wonder, however, if put to the test, would Cisco's UCS beat the best that Dell, HP, or IBM have to offer? If we looked at just what IBM had to offer, how would Cisco's UCS compare with IBM's BladeCenter, Power Systems or System z offerings.

Update:

Todd sent along a note pointing out that he didn't agree with my analysis of Cisco's goals. After reading his post, I think that we're really using different words to get to the same end point. Either way, here is what he had to say:

From: Todd Brannon Sent: Friday, January 27, 2012 6:28 AM

Hi Dan, The comparison is frequently made, given the degree of unification achieved in the system, but Cisco won't claim that the goal with UCS was to recreate the mainframe. The design objective for UCS was to radically reduce complexity and give customers a way to manage across the physical and virtual environments without the burden of integrating an enterprise-class x86 computing platform from the piece parts the industry had offered them to date.

We didn't get around to performance in our conversation but there is a lot to tell there! Since Cisco introduced UCS in 2009 the system has delivered 54 world record benchmark results across the application spectrum. To your question on RISC, last year UCS outperformed a IBM Power 780 on the SPECjEnterprise2010 benchmark. (http://tinyurl.com/74leasw)

Here is a good summary of our performance dominance, particularly in the x86 2 and 4 socket space: http://tinyurl.com/7sb948h

A full account of our benchmarking adventures is cataloged here: http://www.cisco.com/en/US/prod/ps10265/industry_benchmarks.html#~application_performance_reports

Talk soon, Todd Brannon Cisco Data Center Marketing

Editorial standards