BSkyB retunes IT to ready-for-anything datacentre quad

British broadcaster BSkyB runs four resilient, interconnected, energy-efficient and cloud-ready datacentres in the UK. Yet just a couple of years ago, the company was nowhere near possessing a future-proof foundation for all its IT, Riccardo Degli Effetti, head of datacentre operations at BSkyB, told the Datacentres Europe 2014 conference

“Between 1990 and 2005, we had no datacentre strategy and no IT approach,” Effetti said. “We were putting a server pretty much anywhere we could get them to work. But it clearly didn’t work for us.”

The IT team embarked on a datacentre efficiency project in 2010 after deciding to consolidate its datacentres and make them cloud-ready and future-proof. BSkyB has six datacentres – one in Amsterdam, and five in the UK (at Land’s End, Slough, Hemel Hempstead and two in Edinburgh).

“Our four core datacentres are our most resilient and efficient,” Effetti said.

The IT team outlined its main objectives and goals before devising the new datacentre strategy. “We had a few strong objectives,” Effetti explained. “We wanted to save the company some money and for our datacentres to be the foundation of all BSkyB IT. 

“We also wanted a highly flexible, highly available, future-proof facility ready to take on any challenges – because you don’t know which of the emerging technology trends are going to hit your IT.”

Building a datacentre in-house was not an option. We do TV business not datacentre business

Riccardo Degli Effetti

The broadcaster decided to use co-located datacentres for its four core facilities (the Edinburgh pair and the M25 pair (Slough and Hemel Hempstead).

“We understood that building a datacentre in-house was not an option because it wouldn’t be cost-effective to get the scalability and resilience that we want in-house,” Effetti said. “We do TV business not datacentre business.”

The team made a conscious decision not to use one of the big four datacentre/co-location providers (Equinix, Interxion, Digital Realty, Telecity). Instead it opted for smaller third-party providers. 

“We use more than one provider because we have four facilities,” Effetti said. “We wanted attention and good service from our co-lo providers.”

His advice to companies considering co-location service providers is: “Take time and choose your partners carefully and have a good relationship with them. If you consider a partner that is wrong for your needs, your costs will spiral.” 

Today the company runs four resilient and highly-efficient datacentres. Between them, they are equipped with 5,000 servers and host all BSkyB’s mission-critical applications and workloads. Each facility has a power usage efficiency (PUE) rating of between 1.38 and 1.45; the industry average is around 2.0.

“Within our datacentre operations, where we can really make a difference is to make them sustainable and environment-friendly,” said Effetti. “As soon as we started moving servers from old datacentres into new facilities, we saw our power costs going down by 30%.”

In one of the Edinburgh facilities BSkyB uses free-air cooling. “This alone helps us avoid carbon emissions equivalent to 10 flights from Edinburgh to London.”

But the company hasn’t got rid of the other energy-inefficient legacy datacentres. “We are not getting rid of the old datacentres any time soon,” Effetti said. “They have had a lot of investment and we still use them for legacy stuff.”

So what else did the BSkyB IT team do to make its infrastructure more environmentally friendly?

Most downtime incidents in a datacentre are caused by people, pointed out Effetti. Human error in datacentre facilities includes mixing up the Fahrenheit or Celsius setting, accidentally pulling out power cables, or overloading a circuit by plugging in another server.

“We avoid anyone getting inside as much as possible,” Effetti said. The servers are encased in glass boxes, so the IT team can gather data insights through the glass rather than by touching the units.

As for the networking infrastructure, it is all diversely connected so there is no single point of failure. The team also uses datacentre infrastructure management (DCIM) tools to identify and fix inefficiencies rapidly.

“Servers cost money and they have to be implemented carefully and optimised, so we use DCIM tools,” he added.