After reading Compare the Cloud’s piece on data centres in 2020, I’m offering you all my own thoughts on the future. The data centre in six to ten years time will undoubtedly generate power locally but I disagree that it will be based in nuclear fission.
I do agree that the maturity of micro mobile devices and strategies will drive massive growth in data analytics, storage and processing. Also, I doubt that anything can have a clear purpose when we call it the ‘Internet of Things’. Did we learn nothing from prefixing all products and services with ‘e’ 15 years ago!?
Power is the greatest cost to any data centre provider, regardless of global location – so becoming cost competitive and more sustainable requires us to begin to generate and source power differently. After all, the greatest proportional rise in power costs have been due to distribution costs not power production. data centres of the future will witness a complete re-engineering of how they process, manage, connect and store. We will evolve from buying homogenous blocks of capacity; whether it be defined by units of racks or the time it takes to compute.
The data centre will morph into banks of homogenous modular processing, flash and connection concentrations that drive out many of the inefficiencies and under-utilisation of previous generations.
In the future we will purchase capacity subscriptions in the same way that as domestic and commercial consumers we buy power, television and mobile services today. Connectivity will be ubiquitous, one package which connects our lives and all our devices.
Subscription to connectivity packages will provide limitless connectivity regardless of the device and locale to hubs which process, protect and store our unique data. These will drive a transformation within the data centre, a movement to large blocks of centralised processing and memory, which is allocated based on a quality of service and subscription package.
‘But that’s a Mainframe’ I hear you say, well yes, processing will be delivered from banks of clustered devices – be they CPU or GPU, which will be consolidated and optimised with the emphasis on intelligent grid management systems which only invokes more capacity on demand. These will be supported by flash systems delivering shared memory and the first tier of storage to support the ‘always on’ instantaneous experience we are now beginning to demand.
Multiple levels of storage will then determine by touch, data age and retention profiles as to how data should be stored and indexed. All this will be managed by the connectivity strategy which constantly analyses our data streams, usage, and services aligned to the best subscription packages to allocate resources most effectively. The data centre will morph into banks of homogenous modular processing, flash and connection concentrations that drive out many of the inefficiencies and under-utilisation of previous generations. Let’s think of these as Smarter data centres.
So what powers these SDCs? I remember reading about the impending maturity of nuclear fission 20 years ago and the excitement of what it could become; yet taking nuclear fission mainstream remains our greatest challenge. data centres need two attributes; continuous power and connectivity. The source of either is, and will remain only important in terms of unit cost.
Over the years I have seen many data centre projects consider a wide range of local power sources. Whilst nuclear fission will undoubtedly solve a number of issues, regulation and risk will ultimately govern its adoption. Creating smaller local pods of nuclear waste production from data centre fission cells will need a clear strategy; as will managing and transporting fuel and waste to restricted and registered sites.
Fission research led to the creation of nuclear weapons, and building a nuclear capability within a data centre facility will require another level of control and governance which will need political approval.
Given the challenges we have faced already in getting clear, considered and well invested strategies for long term sustainable power generation in the UK by already approved and regulated suppliers, the likelihood of this cascading to local regulated operators in 6 to 10 years is not conceivable even if we are being optimistic. Are we really going to see mainstream global capitals approving nuclear fission cell deployments without the appropriate regulation and risk management frameworks in place?
The data centre of the future will be locally sustainable. We have to address the issue of continuous power in terms of many generation strategies such as biomass, solar, wind or geothermal as many of these require maintenance at some point, as will nuclear fission infrastructures so we either ultimately fall back to national infrastructures which need to be sized for peak demand or we deliver this through temporary or even permanent redundant onsite generation.
The data centre of 2020 will have three main considerations. There will be a need for more space to support the generation strategy. We will be required to change our thinking around long-term capital investment strategies; to support either the refit or implementation of new technologies. The new technologies will trade long-term unit cost reductions for even greater initial investments. Finally, there will need to be greater central and political support.
Without considerations in the realms of investment, policy, governance, and innovation in 2020 we will purely deliver the ‘Internet of Things’, there will be no clear definition, and confusion will continue to reign supreme.