Demand for virtualised datacentres is high, they said, due to the complexity of managing and provisioning physical resources, securing that environment, maximising utilisation of assets, numerous network connections, and the rising costs of facilities and energy usage.
“Power is increasing at a faster rate than the top line revenue of your company,” said John McCool, senior vice president and general manager of Cisco’s datacentre, switching and services group. McCool spoke at the Cisco Live conference in San Francisco.
Virtualisation removes the logical view of an infrastructure from the physical underpinnings, thus making datacentre resources transparent to an application and enabling that application to move, McCool said. Cisco itself was faced with a “$100 million server” issue – it needed the server but didn’t have enough room for it in a current datacentre and faced an expensive build out of a new facility just to house it.
But virtualisation let Cisco pocket that stash. Virtualising its datacentres reduced its cable plant by 4,800 cables, McCool said, made room for 50% more physical servers and increased virtual machine capacity fourfold.
In another example, a New Jersey financial institution opening an office in Bangalore opted to host applications in New Jersey and implement virtual desktops in India with Cisco’s Wide Area Application Services and Application Control Engine products to save money, McCool said.
But virtualisation makes the mobile VM difficult to monitor and track, he said, and thus hard to manage. That’s why Cisco developed VN-Link, software that allows the network to become VM-aware and map policies to a VM as it moves across physical ports.
VN-Link is intended to provide full visibility to VMs for the network administrator and VM management for the systems administrator, says Ed Bugnion, Cisco’s CTO for the server access and virtualisation business unit.
VN-Link is integral to Cisco’s Nexus 2148T fabric extender to provide network interface virtualisation, which provides a “direct, consistent view of VMs and the (datacentre) operational model” by divorcing VMs from their physical interfaces.
This, combined with a unified switching fabric supporting Ethernet and Fibre Channel, simplifies and provides greater and more consistent visibility into all datacentre operations, Bugnion said. The environment can be managed from a network perspective or from a server perspective, he said.
Cisco’s platform for enabling all of this, of course, is the Unified Computing System. UCS is intended to be a single point of datacentre management through its ability to discover, view and configure resources, such as servers, and apply and enforce service profiles on those servers.
“Our strategy (with UCS) is to accelerate virtualisation through increased visibility and control,” Bugnion said. “We’re focused on that part of the datacentre in which the network plays a central role. Management and energy efficiency is not an afterthought.”
To date,Savvis and Thomson Reuters have discussed significant benefits from using or trailing virtualisation and UCS.