Green Data Centres – It’s Easy, Follow the Sun

[#1]

If you are an advocate of solar energy, then one of the biggest debates that you are likely to get into is around the question of Grid Parity. If you carry out some simple sums, then in terms of the cost to build a new power plant, we are already there.  The supporting evidence is the number of new base load solar plants that are springing up, like this one in Arizona. But the phrase that has captured my interest is that of a Solar Continent. Australia is, like much of the US, very well positioned to capture a great deal of energy from the sun. With the falling prices of PhotoVoltaic (PV) panels, it is very likely that in the next ten to twenty years much of the western world will be powered with some form of distributed energy (DE). So how does this affect our data centres?

The problem our data centres have is that they are unable to tap this energy early enough to shed the large running costs now being experienced. Many of the world's data centres are not on a path to long term reliability or sustainability. With the complex business interactions now necessary to complete transactions, the availability of reliable energy is paramount. Consequently, the state of data centres is now seen by industry observers as being parlous; however, as outlined in this Synnovation article, The Future Looks Green, the path to sustainability is not insurmountable.

From a data centre perspective, the major issue with solar energy is the limited number of effective solar hours in a given year. The shortfall in solar availability is normally compensated for with energy storage devices that are presently expensive and to some extent experimental. In reality, the shortfall in energy demand is often taken up by drawing on the centralised grid supply. However, a novel approach to this problem is to consider moving the processing instead. So, on a cloudy day or at night, rather than switching to the grid or storage, we could just move the processing to another data centre. This way, we build a virtual data that is truly "Following the Sun". There would be, of course, a set of "goodness" metrics or broker that would ensure that processing only went to data centres that could provide sustainable energy.

With today's technology, a Distributed Energy and Processing Balancing System would not be hard to build. Virtualisation in terms of server or application images is well established, and with those images shared between the contributing data centres, a set of pointers could provision a standby image in say the West Cost of the US, (or even Africa) very quickly. But how might we move or share the data? Moving large volumes of data, long distances, is of course not viable, however implementing a distributed peer to peer network is. All data centres that are part of the virtual data centre would share the same data (with perhaps some small local caching). Now this is already being done with many peer to peer systems, using technologies such as Distributed Hash Tables and Scalable Distributed Data Structures for example.

In the future, we may need to re-evaluate our data centres as they transform from lumbering consumers of energy to truly green and agile data centres participating in a dynamic balancing of processing capacity and sustainable energy. So, in the future, it will be not about trying to move the energy to the processing, but moving the processing to the best energy source!

[#2]

The conversation I want to hear is around the fact that data centers produce heat - an untapped form of energy. Couldn't water-cooled servers help created the building's needed hot water supply -- or turn turbines -- or such like?

It seems strange that, as a 'best practice', rather than adopting a geo-thermal heat pump approach at data center architecture ( en.wikipedia.org/.../Geothermal_exchange_heat_pump ) we just agree to consume more electricity trying to air condition them.

[#3]

You have raised a very good point. Indeed, many of the concepts are built on the premise of sustainability, which requires a great reduction in energy consumption in order so we can reach the tipping point that makes renewable energy viable.  A strategy that can be adopted was suggested in my article  (referenced above).  In this article, the issues of effiencies techniques of higher efficiency, utilization, virtualization, data centre design, DC conversion etc, are all addressed towards reducing the main load so that the base load can be reduced.  I have developed a roadmap that demonstrates this.