Cisco Updates Unified Computing Servers With Westmere, FEX-Link

New fabric extension architecture rolls out as Cisco refreshes its Nexus, MDS and UCS offerings with new CPUs and capabilities.

By Sean Michael Kerner | Posted Apr 6, 2010
Print ArticleEmail Article
  • Share on Facebook
  • Share on Twitter
  • Share on LinkedIn

In a widespread refresh of its datacenter gear, networking giant Cisco Systems is updating its Unified Computing System (UCS) rack and blade servers with new Intel CPUs and an improved fabric extension architecture designed to simplify network management. Cisco's Nexus and MDS switches are also being updated as part of the effort.

Cisco's (NASDAQ:CSCO) rollout of its second-generation M2-series UCS offerings leverages Intel's new Westmere CPUs to provide additional cores and performance. Cisco had previously talked up the UCS M2 series during Intel's (NASDAQ: INTC) Westmere product launch, though the products themselves debuted only this week this week as part of the official UCS generation-two product update.

There are five new Cisco UCS M2 servers powered by Intel Westmere CPUs, including the B200 M2 and the B250 M2 blade servers, each of which uses a two-socket Intel Xeon 5600 CPU. The B200M2 supports 12 DIMMs while the B250 M2 offers 48 DIMMs. On the rack-mounted server side, Cisco introduced the new C200 M2, a two-socket, 1U server with four disks and 12 DIMMs; the C210 M2, a two-socket, 2U design with 16 disks and 12 DIMMs; and the 2U, two-socket C250 M2 with eight disks and 48 DIMMs.

In addition to the new Intel CPUs, Cisco is updating its fabric extension technology, which also benefits Cisco's Nexus switches.

"Fabric Extender is the name of the product we've been using all along, and the acronym is FEX," Kash Shaikh, senior manager for the data center solutions marketing team at Cisco, told InternetNews.com. "We are introducing a new term, which is FEX-link, as an architecture, as it's the same underlying architecture in the UCS and Nexus product family."

Paul Durzan, director of hardware platform marketing at Cisco, explained to InternetNews.com that when blades first came out, they were typically hooked up to the network through passthrough modules. The industry then moved to a switching architecture that meant network administrators have had to manage switches in the blade chassis in addition to the network switch, which adds to overall management overhead.

The idea with FEX-link is to simplify blade management and to help reduce complexity. Durzan noted that Cisco is placing FEX in the UCS chassis, which is all managed by a Nexus switch and which provides a single management domain from a network perspective.

"We have a nice relationship on the networking side, and with UCS, what you'll see is you'll be able to get up to 160Gb/s per blade of I/O with the FEX-link architecture," Durzan said.

Switch and Software Updates for Improved Network Management

Alongside the new UCS M2 servers, Cisco is updating its line of Nexus 2000 switches, which act as fabric extenders to help reduce management overhead. The new Nexus 2232 is a 10 Gigabit Ethernet fabric extender while the Nexus 2248 is a 100 Mbps/1GbE device.

Also being rolled out as part of its networking refresh in addition to the UCS and Nexus updates, Cisco this week unveiled its new MDS 9148 Multilayer Fabric Switch, which features 48 8Gb Fibre Channel ports. The new switch is targeted at virtual server environments and offers scaling from 16 to 48 ports with on-demand upgrade licenses. Cisco also said that it has added 8Gb Fibre Channel uplinks to the Nexus 5000 Series and UCS 6100 Series interconnects.

In addition to hardware improvements, Cisco is making enhancements on the management side as well. The UCS Manager which handles management of the Cisco servers is now being expanded for developers.

"We've decided to create a Cisco Developer Network where developers can access the SDK to get trained on how to write to the UCS Manager APIs," Durzan said.

UCS Manager is also being expanded with new server service profile capabilities, with enhancements that include new power and cooling policies.

"In the past, you would have had to manually go to a server and put on power and cooling policies," Durzan said. "By putting it on a service profile, you no longer have to manually control your environment in terms of power, and you can automatically associate it with an application."

Cisco has a broader corporate effort called EnergyWise, which aims to help enterprises manage their networking power usage. The new UCS Manager power service profiles, however, are not part of that effort.

"We're talking to them [EnergyWise] and EnergyWise is more on the switching side by we haven't put together a full integration program yet," Durzan said.

Sean Michael Kerner is a senior editor at InternetNews.com, the news service of Internet.com, the network for technology professionals.

Comment and Contribute
(Maximum characters: 1200). You have
characters left.
Get the Latest Scoop with Enterprise Networking Planet Newsletter