Cisco on Thursday unveiled the third generation of its Cisco UCS fabric computing platform centered on new servers based on Intel's Xeon E5-2600 processors.
The new UCS platform also features enhanced unified management technology for increased network and data center scalability, and improved bandwidth and scalability for virtualized environments, said Todd Brannon, Cisco's senior marketing manager for unified computing.
The new third generation Cisco UCS platform builds on the momentum UCS has shown since it was first launched two years ago this month, said John Chambers, chairman and CEO during a Thursday webcast introducing the new technology.
That momentum includes a current UCS server annual run rate of $1.3 billion, and a year-over-year growth rate of 90 percent in the last quarter, Chambers said. "Once again, Cisco is majorly committed to leading in the data center, and committed to leading in the cloud," he said.
Cisco UCS is a converged infrastructure offering which combines Cisco's server and networking technology into a single managed solution. It is the server and networking foundation on which both VCE's Vblock and NetApp's FlexPod converged server-storage-networking solutions are built.
Cisco introduced a new blade server and two new rack mount servers based on Intel's new Xeon E5-2600 processor.
The three servers, expected to ship later this month, include the Cisco UCS B200 M3 half-blade form factor blade server which supports 24 DIMM slots of memory and up to 80 Gbits per second of I/O bandwidth, the 1U UCS C220 MC rack mount server targeting a wide range of business work loads from Web services to distributed databases, and the 2U UCS C240 M3 rack server targeting more storage-intensive workloads including big data and collaboration, Brannon said.
On the management side, Cisco UCS Manager now better manages both blade and rack mount servers in a single domain, Brannon said.
"Last year, we first allowed rack mount servers to connect to the fabric," he said. "But this used up ports quickly. Now we're adding the Nexus Fabric Extender to scale the fabric while reducing costs. So now we can scale to up to 160 rack or blade servers in a single domain."
Sometime during the second half of this year, Cisco will release its Unified Management for Multi UCS Environments, Brannon said. That technology will tie multiple domains together into global compute pools of up to 10,000 servers. "It keeps the current UCS API so that disaster recovery and business continuity can be extended using customers' existing tool sets and existing technology partners," he said.
Also new is the 2204XP chassis which provides options for 80 Gbps and 160 Gbps bandwidth to handle workload bursts, Brannon said. The new chassis includes port channeling to allow load balancing across ports.
Cisco also embedded its existing virtual interface card on its blade servers which resulted in a 30-percent decrease in CPU utilization and a big drop in latency when running virtualized applications, Brannon said.
"This lets customers virtualize applications that before were left on bare metal because of performance constraints," he said. "By allowing more use of virtualization, it increases customer ROI (return on investment)."
As a result, the new servers have an embedded 40-Gbps adaptor which can be carved into up to 256 virtual adaptors. "With the new chassis, total bandwidth can be upgraded to 80 Gbps," Brannon said. "80 Gbps is an absurdly large bandwidth for a single blade."
Cisco also unveiled the Cisco UCS 6296UP Fabric Interconnect which doubles the switching capacity of the UCS fabric to 1.92 Tbits per second to reduce latency.