Over the past year, twin pressures have been building on data centre managers. Natural disasters, human errors and catastrophic equipment failures resulted in some high-profile interruptions to services; data centre managers now face a renewed focus on service availability. At the same time, increasing energy costs and corporate eco-awareness have heralded an era of energy-conscious IT. At first glance, these two goals would appear mutually exclusive.
Historically, IT has guaranteed service provision by building in redundancy. Back-up power supplies, failover servers, even dark data centres (identical facilities kept in a state of readiness in case the main site is lost) have been used to assure that services remain up and running – or that disruption is minimised. Nevertheless, each additional piece of equipment that is added to bolster service assurance comes with an energy overhead. How then, can a business be both energy efficient and resilient?
Turn away from the dark side
For a start, dark data centres are a non-starter. As Forrester analyst Stephanie Balaouras points out, the combination of cost pressures and the shortage of prime data centre real estate have already convinced most businesses that these facilities are no longer viable. “Given the necessary investment, an alternate data centre simply can’t remain idle waiting for some disaster to occur,” she says.
But even within operational data centres, inefficiencies are deeply ingrained. Neil Rasmussen, CTO at data centre infrastructure company APC, bristles with irritation at the suggestion that back-up power supplies are a major cause of inefficiency: some of the older models were inefficient, he concedes, but not the latest generation. Furthermore, buyers are increasingly being supplied with efficiency ratings for data centre equipment – invaluable in helping them balance efficiency and resilience, he insists.
Even in those rare cases where organisations can kit out a data centre with entirely energy-efficient equipment, there is still a high risk that the facility will be skewed by a focus on resilience. Typically, data centre managers “over-provision” when it comes to their data centre, suggests Vic Smith, enterprise technologist at hardware-maker Dell’s data centre infrastructure group. “Given a typical 5,000-plus server data centre, there will be maybe 100 to 200 systems that are absolutely vital. But all of the systems will be lumped into the same data centre,” he says. “It would make more sense to take a compartmental approach, and really focus on building resilience where it is most needed.”
For others, such as Philip Cheek, director of business development at data centre services group Inifinity, balancing energy efficiency and resilience also means businesses need to consider where they are today. In many enterprises, the IT infrastructure has grown haphazardly as the business has grown. As a result, it is common for organisations to have multiple facilities that might be regarded as data centres – from repurposed broom cupboards to fully specified server rooms.
Consequently, Infinity regularly sees clients looking to consolidate this sprawl into perhaps just three facilities. “That in itself provides the opportunity to be more energy efficient,” says Cheek.
Some of IT’s big thinkers, such as pundit Nicholas Carr, believe the answer to this conundrum lies in the cloud – as IT service delivery moves towards a utility model, reliability and efficiency is achieved through scale.
Google is frequently cited as the ideal. It has built several mammoth server-packed hangers, and it is thought that the company has achieved significant energy savings. Its new
But interruptions to service are still possible. In February 2008, users of Amazon’s S3 storage-on-demand service had first-hand experience of the fallibility of cloud computing services.
In the short term, virtualisation is a less traumatic solution for those reluctant to embrace cloud computing. It is already helping many companies deliver a resilient commodity server infrastructure, as applications run on virtual machines can be easily moved and are less reliant on the uptime of individual components of the physical infrastructure. Server virtualisation has also helped drive up utilisation rates – giving additional energy efficiency benefits.
With storage virtualisation now beginning to inch into the enterprise data centre, it promises to bring similar benefits to the storage element of the data centre. And while this is still some distance from the cloud computing promise of unlimited processing power delivered as a service, it will at least tick the boxes for resilience and energy efficiency.
Energising the Green IT debate Two new reports on power consumption highlight the urgent need for greater energy efficiency in IT