A blueprint for greener cloud computing

If you’ve logged onto Facebook or watched Netflix, you’ve experienced the benefit of cloud computing. If you’ve used Dropbox or Adobe’s Creative Suite, you’ve seen how cloud computing is triggering change in the workplace.

Instead of purchasing a high-end computer to run newer programs, you just need an Internet connection. Cloud computing is an extension of your existing computer or mobile device, enabling you to perform demanding computing tasks without placing extra pressure on your device. It is no longer ‘nice to have’ – cloud computing is an essential part of any contemporary organisation and day-to-day life.

Cloud computing reflects a pay-as-you-go model that places information, data storage and processing power at your fingertips. Businesses no longer need to invest heavily in IT infrastructure or agonise over system maintenance, but can instead leverage large-scale computing resources that data centres pool together.

The smaller services businesses need are mapped to physical computers sitting in the data centres, creating an arguably greener model of IT.

More recently this model has been extended beyond the business world and into developing countries, who can tap into existing applications rather than investing in costly infrastructure, and public services, where hospitals can easily access clinical information to improve care and save costs.

While hardware costs are decreasing, and constructing data centres to support cloud computing is more feasible than before, operating costs are rising. ‘Always on,’ their energy usage is comparable to a small town and represents more than three quarters of total running costs.

The carbon footprint of data centres is bigger than the Netherlands.

While it is no surprise to learn data centres’ carbon footprint was bigger than the Netherlands, it is concerning to hear their electricity consumption will reach over 140 billion kilowatts by 2020 in the US alone. Data centres promise a greener approach to computing, but cloud computing has not yet harnessed their full potential.

Many data centres are designed to handle peak periods, but squander energy in between these times. The cloud has not yet optimised the power of data centres, but my research seeks to devise a new approach to allocating work through the cloud, creating a new generation of computing to reduce environmental impact and produce better services for consumers.

Our first approach to cloud optimisation involves a consideration of user service requirements. One consumer may want a service delivered right away, while another user might want their service delivered in an hour. For immediate service, I may charge $2. But if the customer can wait, the charge may only be 50 cents. Consumers can save more money if they accept a longer delivery time, so there is an incentive to choose the delayed option.

But whether the service is delivered in ten minutes or an hour, it still requires the same amount of work from me. So how can I offer one at a cheaper rate?

If I want to travel from the University of Melbourne to Melbourne Central, there are a few different options I can choose from. I can take a taxi, which could cost $10 and take five minutes. I could catch a tram, which would cost $3.76 and take 10 minutes. Or I could walk. While this is free, it takes a lot more time. In this scenario we’re thinking about the cost to the consumer, the time taken, and the amount of power the transport requires.

We can allocate tasks with cloud computing in a similar manner. If we use higher frequencies, we can perform tasks faster but will consume a lot more energy.

If we use lower frequencies, tasks will take longer but consume less energy.

As shorter timeframes are more expensive and longer timeframes are cheaper, there is an incentive for both service provider and consumer to favour the option that best suits their priorities. Therefore, we can integrate the time in which a consumer wants a service delivered, or quality service requirements, into specialised algorithms to optimise cloud computing.

We can take this model further by inserting computer availability, capability and performance into the algorithm. From here we can automatically identify which computers are overbooked and which are underbooked. With this approach we can harness the full potential of data centres, enjoying the benefits these enormous centres provide while at the same time shrinking our carbon footprint.

In the next two decades, service-oriented computing will play an important role in shaping business. But we need cloud infrastructure to host applications and deliver services that will meet users’ quality of service requirements. Only with a novel approach, in partnership with the Australian Research Council, can we take cloud further and generate green technology to minimise environmental impact.

Professor Rajkumar Buyya from the University of Melbourne is Chief Investigator of the research project ‘Algorithms and Software Systems for Management of Software-Defined Clouds’, which has been funded by the Australian Research Council.

Banner image: Flickr