Like many multinationals, BT is sitting on some extremely expensive data centre real estate.

The telecoms services giant runs 23 data centres in the UK alone – with half of its 22,000 square metres occupied by its own systems and half run on behalf of customers. It also has another 35,000 square metres spread across 60 data centres elsewhere in the world, mostly hosting customers’ systems.

Unsurprisingly, not all of that data centre estate is state-of the-art, or indeed exploited to its full potential. Much of it is mainframe-oriented, complex, under-utilised and hard to scale. It also costs too much to run.

Over the next two years, BT plans to redraw that picture radically. Setting a target of removing £40 million in annual costs from its data centre operations, the company has embarked on a transformation programme designed to deliver industrial-strength, commercial-grade data centre services – with known unit costs, dramatically improved utilisation of infrastructure and data centre space, applications independent of any host, and services that are responsive to change and that can be scaled vertically and horizontally.

As that suggests, the consolidation and migration of business services is on a massive scale, involving systems, storage and applications. According to the head of data centre transformation, Finlay MacLeod, to cope with that BT is employing a specially developed ‘Smart consolidation’ methodology that will allow it to swap out vast quantities of “old tin, for more efficient tin – blades, virtualisation, containers, and so on”. In many cases, that will see mainframes being replaced by Windows, Unix and commodity storage and network systems.

But there is no silver bullet here, says MacLeod – certainly not blade servers. The standard approach by many other companies has been to consolidate ‘as is’ on blades, he says, but this fails to deliver maximum benefit: “Don’t put everything on blades; that does not buy you much except high energy costs.”

A level up from the infrastructure, a huge applications consolidation undertaking is occurring. That will involve breaking the link between applications and the host they currently run on (through virtualisation); reduction of the number of instances of existing applications (which should lead to licence cost reductions); the standardisation on some applications and the retirement of others; and the removal of ‘orphaned’ applications that have few, if any, users.

To ascertain the scope of the task, BT has been using ‘discovery tools’ to help its teams understand “just what it has in the engine room”.

“In migration planning, the key is an understanding of application configuration,” says Macleod. The Smart methodology reduces risk by understanding the linkages and dependencies between applications and systems, turning a “craft workshop for each application system into a batch-flow production line.” However, not all applications will be making the journey: BT intends to take 147 applications out of the infrastructure.

The programme is not only about more efficient use of infrastructure. BT wants more and better space. “We are upgrading strategic data centres and releasing valuable space for reuse by the business,” says MacLeod. 

How far can the consolidation go? The company has set a consolidation target of 20:1, and given the leaps in the performance and concentration of processing power in recent years – plus the elimination of redundant kit – “in large measure, that’s achievable,” he says.

The pressure does not stop there, though: the programme has to be achieved while maintaining unquestionable systems reliability. “Consolidation without service interruption is key,” says MacLeod. After all, BT runs the 999 emergency call handling service and NHS projects like ‘The Spine’, the national database of information about patients’ health and care that forms the core of the NHS Care Records.

But given the degree of planning and the care it is taking in the execution of its consolidation, BT is, perhaps justifiably, confident that the project will steer clear of the critical list.