Carphone Warehouse's IT department has cut the time it
takes to provision servers to ad hoc business needs from six weeks
to thirty minutes usingvirtualisationandautomation technology.
Configuring the right amount of virtual servers to meet
application demands automatically, rather than manually configuring
physical servers, underpins the business's long-term strategy to
offer new broadband services such as voice, video and data to
customers reliably and quickly, said Steve Gall, lead datacenter
architect.
He said that early trials of automatic server configuration
represented approximately half a million pounds in labour savings
and better use of server capacity that had previously been
underused.
"The phone industry is fast moving and our IT department have
found it difficult to keep up with the demands of the business in
the past. We're dealing with 360 terrabytes of information now, and
by using automation we've increased the utilisation of our server
capacity from 17% to 60% which has reduced our operating expenses
by 30%," said Gall.
The company uses a combination of IBM P-series, Sun Solaris,
Linux and Wintel servers in its data centres and has deployed
software from Enigmatec for provisioning. Carphone Warehouse uses
the
Information Technology Infrastructure Library (ITIL) version 2
framework for establishing service level agreements (SLAs) of
applications with the business. It chose Enigmatec's software
because it shows how making changes in server provisioning in one
application will affect the service levels in another.
"If you are using automation, then the ITIL framework is
essential. You cannot negotiate an SLA for a piece of hardware with
the business, since that hardware might be used by several
different applications in a virtual environment. Instead, you need
to negotiate the delivery of a service or business application as
the basis for your SLAs," he said.
Gall is currently overseeing a pilot project where the spare
processing capacity on the company's development and test servers
can be re-allocated to support disaster recovery and automatic
fail-over - reducing the need to buy additional servers
specifically for recovery. This will finish in November will be
used as a company-wide model for providing IT services starting
next year.