Christopher Crowhurst is a selfproclaimed tree hugger. He would rather think that the data center efficiency achieved equates to a particular number of Hummers removed from the roads than think about efficiency in terms of the bottom line.

To his disappointment, his role as vice president of strategic technology at information behemoth Thomson Reuters forces him to consider the latter more often than the former.

Crowhurst’s division has recently undertaken a huge programme to virtualise as much of the firm’s IT as possible. But it was not his love for trees that spurred the initiative.

“I’m doing this project because our infrastructure keeps growing at an alarming rate,” Crowhurst said. “We deploy servers at a rate far greater than we make money.”

In the past five years Thomson Reuters experienced an annual growth rate of more than 8%. So why would Crowhurst worry about the cost of IT infrastructure exceeding that figure?

Between 2003 and 2008 the company’s physical storage footprint increased almost eight-fold, currently growing at about 50% per year and forcing Crowhurst to deploy more than six petabytes of storage in Thomson Reuters’ Eagan, Minnesota, data center this year alone.

Within the same period the facility’s network I/O throughput increased six-fold and the number of processors grew 425%.

One of the consequences was the facility’s power consumption growing by 20-30% per year. If Crowhurst continued to do business as usual, the power grid in Eagan would soon be unable to satisfy Thomson Reuters’ demand and the firm would need to pay to upgrade it.

Using the traditional expansion model, the company would have to build a new 2.5MW data center module every year for its IT capacity to keep up with its business growth, spending $500m over the next decade.

“Our company turned to us and said ‘you just can’t spend that money’,” Crowhurst recalled. Buying storage and servers was the division’s largest capital expense, so it turned to examining the way it bought and used storage.

While common practice was to buy tier-1 disks, statistics revealed that only 30% of the storage allocated was used. Newly allocated storage capacity also sat unused for the first six months.

Deploying thin provisioning, Crowhurst’s team was able to reclaim half of the allocated storage, leaving a small cap for growth. Because a lot of the information stored was not mission-critical the team decided to also move a lot of the data that was rarely accessed from tier-1 disks to “cheap and deep” Sata drives.

Virtualisation was the answer to moving storage around according to need. The team only went after the shared-storage environment – a big portion of the total – leaving high-performing applications with dedicated arrays and files.

Crowhurst expects storage virtualisation to save about $8m per year – money he will invest in server virtualisation.

This initiative has proven to be a harder sell, given that generations of what Crowhurst calls “server huggers” were raised by the traditional IT capacity mindset.

“We have server huggers everywhere in our company. Everyone wants their own: ‘I want my own development box; I want my own QA box; I want my own pre-production box and, guess what, I want my own production boxes. And there is no way I’m sharing with any of you guys because I’m special; I’m important; my application needs all that performance.”

Crowhurst’s division has been trying to introduce server virtualisation for years and was met with much resistance. The division assembled a team whose members worked with the various business units to identify applications that could be virtualised and found that about 36% of the servers were candidates.

“It was like pulling teeth. A lot of people fought the idea of virtualisation, so we went after the non-production environments.”

The team put some applications, such as websites in the category of ‘golden geese’. “We were not going to touch those until we knew what we were doing.”

As the division was not in the position of requiring the business units to virtualise, it had to create incentives. The best incentive (surprise) was paying for the process out of the division’s budget – even creating a slogan for the campaign: “P to V is free”.

The team’s target was to virtualise 1,000 servers through growth and this year they will hit 700-800 growth servers. To offset the shortcoming and meet the energy target the firm is virtualising existing physical servers.

There needs to be a change in the way firms think about their IT needs, Crowhurst said. “We aren’t yet at a point in our governance model where we can mandate that ‘virtualise first’ is the mindset. We want to get to a point where we can dictate that virtualisation is the right way to go for everything.”

This article first appeared in DatacenterDynamicsFOCUS magazine. Click here to register to receive a print or digital edition.