LAS VEGAS – Hoping to position itself as the go-to vendor for all-in-one cloud infrastructure, Cisco Systems launched...
a Unified Computing System (UCS) upgrade at Cisco Live 2011 this week that will bring greater speed and capacity to switching and server I/O and promises a deeper level of automation, the holy grail of the cloud computing network.
Data center infrastructure – and more specifically network fabric for the cloud – has been at the heart of the battle between Cisco and HP since Cisco launched its UCS servers three years ago. This week, Cisco boasted that new IDC research ranks it as number two in the U.S. x86 blade server market behind HP and number three wordwide.
With this UCS upgrade, Cisco says it will simplify management of complex virtualized environments and enable easier automated provisioning in UCS environments even as they become more complex for multi-tenant networks that support the cloud. The announcement includes a new Fabric Interconnect, a new Virtual Interface Card (VIC) with dual 40 GbE adapters, a new Fabric Extender I/O Module and upgraded UCS management software.
The new fabric extender will double the bandwidth into the blade chassis while the fabric interconnect will double the system's switching capacity to 1 Tbps. The interconnect also allows administrators to assign any port in the system to Fibre Channel, Ethernet or native Fibre Channel over Ethernet (FCoE) for converged networking. Previously UCS managers were limited to 16 ports for Fibre Channel and had to buy expensive uplink modules to make them work.
“Now they don't need to predefine which ports will be Fibre Channel or which will be Ethernet,” said Soni Jiandani, senior vice president of Cisco's server business.
The VIC has a dual 40GbE interface and enables up to 256 virtual NICs – double what was previously available. That quadruples bandwidth to the server, which Jiandani says “reduces end-to-end latency by as much as 40%.”
Perhaps more important than the new speeds and feeds is the ability to manage the entire system from one central GUI that lets data center managers apply policy to specific applications or groups of virtual machines (VMs) from storage all the way down to a specific VM.
Data center managers can use UCS Manager 2.0 to define service profiles for specific applications or groups of virtual machines so that access and provisioning policies can be applied appropriately. As a result, bare metal applications and virtual applications can run in the same system without overlap and according to their own access and provisioning policies, Jiandani explained.
UCS Manager 2.0 is now integrated with VMware vCenter and has open APIs so that IT shops can integrate other management and provisioning tools as they need them. The ability to manage groups of virtual machines and applications separately through one interface lends itself to automated provisioning and moves away from the idea that data center folks need to slow down the process of provisioning by requesting VLANs or storage access from other teams, for example.
UCS Manager 2.0 also enables IT shops to centrally manage UCS Express blades that sit in Integrated Services Routers (ISRs) distributed across branch offices. These ISR-based blades could boost the local performance of technologies like Virtual Desktop Infrastructure, with desktop applications situated closer to users in the branch office, but still controlled and provisioned from the data center, explained Dave Frampton, vice president of Cisco's application delivery business unit.
Travelport.com, an early adopter of the UCS upgrade, improved how it rapidly provisioned servers during bursts of traffic that are common in its high transaction consumer business.
“Our challenge was that rapid server growth as an online travel company continued to stress our infrastructure,” said Steven Senecal, manager of global server engineering for Travelport, adding that the company's original workhorse servers were not able to meet the transaction needs or be agile enough to grow on demand. With the UCS upgrade, Travelport was able to provision 190 servers with assigned service profiles in three business days.