Cisco plans to rewrite the datacentre design book with a
suite of products and a strategy to provide what it calls a unified
computing system. Experts believe its approach will change the way
servers, storage and networks are managed in future generations of
datacentres.
John Chambers, chairman and chief executive office at the
network equipment giant, says it has taken a
different approach to virtualisation, compared with the rest of
the industry, which will help to lower IT costs.
"We want to allow the world to use IT in new ways. The network
is at the heart of the datacentre. The unified computing system
will reduce the amount of infrastructure datacentres need by a
minimum of 30%," he says.
Virtual building block
Cisco's design philosophy is based on the premise that previous
generations of datacentre equipment are unable to cope with the
demands of a modern, virtualised environment.
"The virtual machine has become the new atomic building block of
the datacentre, creating new challenges and opportunities with the
potential to transform the computing environment and deliver
significant benefits," says Mario Mazzola, senior vice-president of
the server access and virtualisation business unit at Cisco.
Cisco's approach relies on computing, network, storage access
and virtualisation products designed into a scalable, modular
architecture, which the company says can be managed as a single
system.
The servers will be based on Nehalem, the next edition of the
Intel Xeon processor, storage will be available over an ethernet
network link, fibre channel, fibre channel over ethernet or iSCSI,
and centralised management will be available from a single console
called Cisco UCS Manager.
Shape of things to come
Forrester Research believes IT directors should look at Cisco's
plans seriously because the Cisco UCS is the shape of servers to
come.
"Something like the Cisco UCS is a logical outcome when you
start designing servers with a clean sheet of paper," according to
Forrester analyst Galen Schrenk, writing on Cisco's strategy.
"The datacentre has become more dynamic. We want to be able to
move virtualised applications around for availability or efficiency
reasons, but older servers make it hard to keep track of the
relationship between logical and physical resources."
Managing data movement
To do this users will need to manage how the data moves between
datacentre components. Experts agree that the efficiency of
virtualisation in the datacentre is limited by the input and output
(I/O) connections between the severs, storage area network and
local and wide area networks.
Schrenk says in an ideal setting, users will want to control I/O
all the way from the datacentre switch to the rack, chassis,
physical adapter, virtual switch, and finally the virtual adapter
with the same admin command and common system admin policies.
Writing on the analyst website
IT-Director.com, Dana Gardner, principal analyst at Interarbor
Solutions, says, "We are seeing a whole new packaging of the modern
datacentre in a way that may very well change the market. It is
bold, and it is risky. One-stop shopping for datacentres has been
only a goal, never fully realised. In fact, many enterprises
probably do not want any one supplier to have such control,
especially when standards are in short supply. But they need lower
costs and lower complexity."