With competition rapidly increasing in the unified data center space, Cisco Systems officials are trying to separate Cisco's offerings from those of its rivals and answer criticisms about Cisco's Unified Computing System platform.
During an hour-long Webcast April 16, two Cisco executives outlined the innovation and cost savings that have gone into the UCS initiative. Soni Jiandani, vice president of marketing for Cisco's Server Access and Virtualization Group, and David Lawler, product marketing manager for the same business unit, pushed aside rivals' assertions that the Cisco plan is too expensive, doesn't scale and requires a drastic overhaul of the data center.
"The system does scale," Lawler said at one point during the event. "That was the way this was architected from Day One."
Earlier in the day, Cisco announced that it was partnering with NetApp to offer solutions based on the UCS and NetApp's Unified Storage Architecture.
Top technology vendors, including Cisco, Hewlett-Packard and Sun Microsystems, are rolling out solutions designed to unify the disparate server, storage and networking tiers in traditional data centers as businesses look for greater agility and cost efficiency from their computing facilities. Fueling much of the trend is the increased use of virtualization technology, which can help cut some costs but also adds complexity to the data center environment.
Cisco March 16 rolled out its UCS plan, which centered around its own branded blade servers powered by Intel's new Xeon 5500 series chips, aka "Nehalem EP," innovations that the company has introduced, and partnerships with the likes of VMware, EMC, Intel and QLogic.
Jiandani said Cisco's ideas for UCS were driven in large part by the demands of customers, who were looking for ways to improve data center performance-including the ability to scale-while reducing expenses.
"We wanted to embrace innovation while driving down costs," she said.
Key to the Cisco innovations was the creation of the Cisco Memory Extension technology, which works with the new Intel "Nehalem" architecture to optimize memory capabilities within blade servers, which is important in virtualized environments, Lawler said. A problem was that while the CPUs were increasing in power, memory was not keeping up, which limited performance gains.
"Customers were telling us they were running out of memory before they were running out of CPU [capacity]," he said. "They were underutilizing the CPU. Intel had gone from a single core to quad-cores in four years, but the memory was not keeping up."