While IT is often an enabler of decarbonisation efforts, our industry is responsible for 2 - 3% of global emissions. This is a challenge for software, architecture cloud computing and physical infrastructure – so we need to employ a combination of solutions and green technologies.
I was asked to speak on sustainability at a conference in Ireland last month. I planned to draw on my knowledge of how data and AI were helping to reduce emissions through smart spaces, smart places and other ‘smart’ solutions. But then I saw the distinguished line-up of speakers focused on buildings, transport and energy. So, for variety I turned my attention instead to my own industry and sustainable IT. An advantage of being part of the NTT family is that our expertise cuts across the problem space.
Like every other sector, we actually face two challenges: reducing emissions and adapting to the effects of climate change.
The IT industry is not immune to the effects of natural disasters caused by climate change. As heatwaves increase, cooling systems designed for the electronics in data centres and telecoms networks risk malfunctioning. This entails infrastructure failure or triggers protective shutdowns. Greater incidences of coastal or fluvial flooding impairs telecoms equipment – frequently located underground. High winds damage telegraph poles and masts - Hurricane Sandy reduced network coverage by 25% in places. Weather-induced power outages have an indirect effect where backup power is insufficient, or generator fuel is stranded in flooded basements. 45% of data centres surveyed have experienced service-threatening weather events 1.
Another part of the problem is the chronic effects of climate change on the supply chain which prevents equipment being replaced or upgraded, causing capacity issues.
The impact of these risks include interruption to or degraded quality of service, repair costs and safety risks for workers maintaining equipment.
The key then is resiliency, which can be aided by public cloud allowing workloads to move between data centres. Some readers will be aware that a secondary risk to this is capacity issues at ‘fail-over’ data centres if many workloads are moved simultaneously. Spreading risk across multiple data centres and geographies, and reserving compute power in a fallback location is a mitigation. However, this is wasteful of resources.
I believe we will eventually see application architectures and service designs that can be deliberately degraded when there are capacity issues – in the same way that smart grids might power-down less critical functions in the event of power shortages. I never truly understood smart grids until I started to grasp the variability in power supply of renewables. This is a bit of a live issue for the country where I was speaking – Ireland’s booming data centre industry is now putting a strain on the national grid at peak times 2.
Compared to the other industries at the conference, in IT we can iterate our software products – and their subcomponents - much faster, which I think is a key reason why our industry is able to make the biggest and fastest strides in decarbonisation. But ICT as a sector still contributes 2-3% of global emissions. Some of that is right on our doorstep at NTT because we are one of the largest operators of data centres in the world, processing more than 50% of global internet traffic. We’re also one of the world’s largest systems integrators so we have a responsibility to produce innovative green technologies such as carbon efficient software and architectures that make more of the data centre capacity we have.
Use of public cloud for application workloads
Migrating workloads to public cloud reduces emissions on average by 87% 3. That’s primarily because compute capacity can be much more efficiently and effectively shared, and because at scale, cooling can be done in a more energy efficient way. I’ll come back to efficient cooling later.
Use of edge computing
If more compute workloads can be done at the ‘edge’ (e.g. coupled with low powered smart sensors), energy can be saved in data transmission and in energy consumption inside data centres. This complements other advantages of edge computing such as data privacy and speed of response.
This is an important part of what we do at NTT DATA. For this discussion it suffices to say that energy, cost and carbon reduction are often but not always synonymous. Typical savings of 40% in cost translate to a similar carbon reduction, but we are actively working with customers on putting more accuracy around these numbers to re-enforce the double whammy of being more efficient with cloud resources. Will the hyperscalers alone solve the problem by achieving Net Zero? We’re not close enough for complacency. Even for Microsoft that’s eight years away; AWS and Google have more near-term commitments but to renewable power, not zero-emissions computing.
As members of the Green Software Foundation, we look at two areas: software carbon footprint measurement and optimising software design. For measurement there is a standard metric my colleague Gadhu Sundaram has been involved in developing along with members of the Foundation called Software Carbon Intensity 4 – essentially a rate of carbon emissions per functional unit of a software system. This is useful because it provides an efficiency measure which can be the basis of comparison between two solutions. Eventually, the hyperscalers will make the measurement aspects of this straightforward by providing detailed carbon instrumentation. Microsoft seem closest with the Emissions Impact Dashboard, but the grain of information is not there today.
The other key project of the Green Software Foundation is the Carbon Aware SDK 5 which allows workloads to be run at a time or location which the energy grid is most likely to be running on renewables. This could be used for workloads that I worry about in Data and Intelligence such as batch data processing and AI model training. This is complemented by a set of Green Software patterns which are more a work in progress. Front-end examples include using less multimedia, using HTML rather than PDF, using the Accelerated Mobile Pages framework, relying more on static content on content delivery networks, choosing more efficient graphics formats, implementing dark mode UIs and so on.
In the same way that car manufacturers worry both about vehicle emissions in life and the manufacturing process, we need to consider the carbon cost of developing, iterating and maintaining software. This is where speed of change is a double-edged sword: iterating software allows you to bring new features and adapt to user needs but also implies an ongoing overhead. Our application maintenance teams ensure services continue to maintain high up-times, security and resilience – but this still contributes to the emissions of running platforms. As NTT DATA UK&I, we will be bringing forward plans to measure the carbon footprint of software delivery as part of a new set of Net Zero commitments in early 2023. I’m as excited to be working on our internal story as I am to be working with customers on these challenges.
Data centre design and operation
The final piece of the jigsaw is how we adapt existing data centres to cut their carbon footprint and design new ones to be low carbon – literally – from the ground up. NTT DATA and NTT Ltd have recently merged so it will be great to collaborate more closely with Ltd colleagues like Vlad Marinov on how we can bring the strengths of both companies to the problem. In the meantime, NTT Ltd have set themselves some ambitious goals - for example to be net zero across operations by 2030.
Getting there starts with data baseline - understanding what equipment is running in data centres, its power draw, embodied emissions and end of life material recovery. This is a complex task because vendors are re-evaluating their metrics and not all providing a consistent dataset. Once that asset information is known, this needs to be coupled with detailed energy consumption and information about the renewables mix in different data centres or parts of the data centre. Sourcing renewable energy is a priority and for embodied carbon, the data baseline drives scenario planning around what infrastructure to retire, what to purchase and from whom.
When planning new datacentres we can use more radical approaches – for example our new Munich data centre is close to ground water for cooling, saving 30% in energy bills. The new London 1 data centre relies on water cooling rather than chiller units making it one of the most efficient data centre designs ever. In the further distance, NTT is developing a suite of technologies called IOWN that use light to conduct information not just across wide area networks but right into a chip. One of the many benefits of this is very low power consumption circuits. The benefits of this and similar innovations are baked into our decarbonisation projections.
Speaking and blogging is one thing: direct action is another, but now I’m starting to see real change on the ground. It’s an exciting time to be working in sustainable IT – helping customers to drive down their own emissions with new digital solutions while keeping an eye on our own impact in the process.
 The gathering storm: Climate change and data center resiliency. Seattle, USA: Uptime Institute.
 Masanet, E., Shehabi, A., Ramakrishnan, L., Liang, J., Ma, X., Walker, B., Hendrix, V., and Mantha, P. 2013. The Energy Efficiency Potential of Cloud-Based Software: A U.S. Case Study. Berkeley, California: Lawrence Berkeley National Laboratory.