To improve IT energy efficiency and sustainability, IT leaders are often advised to keep hardware in use for as long as possible. This is because a significant proportion of the hardware’s lifetime carbon footprint is consumed prior to onsite deployment.
Logically, extending the life of hardware reduces its lifetime carbon footprint, but according to Gartner analyst Annette Zimmerman, there are two issues IT leaders need to weigh up. First is the security risk that organisations face if they extend the life of IT equipment. Manufacturers may only provide hotfixes and patches for a certain period of time, after which the hardware is no longer able to run the latest software, making it potentially vulnerable to cyber attacks. The other issue with extending the life of IT equipment is that it may reduce productivity.
There is a huge amount of effort across the IT sector to encourage IT departments to update older equipment, taking advantage of Moore’s Law to provide more computational power along with cheaper and faster storage and networking, which software providers can make use of to develop feature-rich applications. For instance, Microsoft recently unveiled Copilot, which comprises a number of artificial intelligence (AI)-powered enhancements within the Office productivity suite and Power tools built on ChatGPT.
Newer hardware may also be more energy-efficient. A modern graphics processing unit (GPU) uses more electricity than an older model, but it can run AI workloads much quicker, which means the overall energy required to complete a particular task may be significantly lower. However, this has to be offset against the embedded CO2 associated with its manufacturing, shipping and disposal.
The challenge of AI efficiency
To train and run a deep learning model in production has a significant environmental impact, warns Jason Knight, co-founder of OctoML.
For instance, referencing analysis from the University of Massachusetts Amherst from 2019, Knight says the research found that the process of training large artificial intelligence (AI) models can emit more than 626,000 lb of CO2 – nearly five times the lifetime emissions of the average car.
According to Knight, achieving AI sustainability requires a combination of datacentre efficiencies, adoption of lightweight machine learning (ML) architectures and significant performance gains from the hardware itself.
“Using specialised chips whose architecture and circuits are designed to handle AI and ML workloads can improve performance and energy efficiency by two to five times,” he says. “Specialised artificial intelligence and machine learning chips work best on specific numeric data types, model types and sparsity patterns.”
Growing focus on sustainability
At this year’s Mobile World Congress in Barcelona, Zimmerman notes that one of the hot topics was sustainability and the circular economy for IT equipment. “When people talk about circularity, the first thought is often just recycling, but it’s so much more than recycling. It’s about reuse, remanufacture and refurbishment,” she says.
HMD Global, which manufactures Nokia phones, was one of the companies whose actions caught her attention at MWC. In collaboration with iFixit, HMD Global has introduced a repair programme. “You can go in as a consumer and say, ‘I would like to extend the lifetime of my device’, and they help you to fix the device yourself. But if you don’t want to do that, you can pay for the repair service, which seems to be at a sensible price.”
Annette Zimmerman, Gartner
This idea of reuse and remanufacture is not limited to consumers. Gartner has begun to see some enterprise IT buyers considering remanufactured hardware.
Last year, for example, Royal Mint signed a contract with Circular Computing for remanufactured laptop PCs. Circular Computing’s circular remanufacturing process produces what the company describes as “second-life laptops”, which meet a new BSI Kitemark scheme that certifies devices as equal to, or better than, new devices.
The BSI’s remanufactured and reconditioned Kitemark scheme aims to verify the processes used for the remanufacture of products to show clients that best practice has been followed. The BSI Kitemark covers the process of returning a used product to at least to its original performance, with a warranty that is equivalent to, or better than, that of a newly manufactured product.
Following a successful trial, the agreed partnership will see the Royal Mint use Circular Computing’s carbon-neutral, remanufactured Lenovo ThinkPad T480 units, designed to meet the coin maker’s specific IT needs while allowing for a flexible procurement approach.
While this shows what is possible, Shane Herath and José Gámez-Cersosimo of the Eco-Friendly Web Alliance argue that IT procurement needs to adapt to support the circular economy. “Organisations should also look to revamp their procurement processes to align with the principles of the circular economy, as the linear economic model of IT consumption and usage is not sustainable. The existing approach to IT procurement must change to reduce the impact that our current throw-away culture is having on the environment,” they wrote in a blog post.
Herath and Gámez-Cersosimo recommend that current IT procurement processes should be shaped by circular economy principles, promoting durability of infrastructure and devices, with emphasis on reusing, remanufacturing and recycling to keep IT resources, components and materials circulating in the economy. In the blog, they urge the IT sector to embrace “right to repair” and explore ways to end the planned obsolescence of technological devices and extend their lifecycle.
Shane Herath and José Gámez-Cersosimo, Eco-Friendly Web Alliance
Cutting datacentre emissions
Beyond end-user computing and mobile devices, there is plenty that can be done to address sustainability in datacentre computing.
For instance, reducing the number of physical boxes has a role to play in reducing a datacentre’s carbon footprint. Rob Tribe, vice-president of systems engineering at Nutanix, says hyper-converged infrastructure (HCI) lowers the number of hardware components needed to run a workload, because it converges separate boxes of servers, storage and compute to eliminate an entire storage area network (SAN).
“This will have an immediate, positive impact in lowering carbon emissions, and significantly reduces over-provisioning by offering bite-sized consumption,” he says. “HCI also enables higher automation and increased utilisation.”
With careful management and optimisation, virtualisation, containerisation, software-defined storage and networking, and public clouds, generally offer a way to maximise the utilisation of physical hardware. By sharing services over a network among multiple users, the environmental impact of purchasing and running on-premise IT hardware is decreased.
According to research from Accenture, public cloud migrations can reduce CO2 emissions by up to 59 million tonnes per year, which is the equivalent of having 22 million fewer cars on the road. Clearly, if workloads are left running, they will continue to consume IT resources. But monitoring usage quickly identifies when and where IT resources are being consumed and which workloads can be throttled back or switched off. This reduces the need to power additional hardware when adding new workloads.
IT departments are also able to extend the useful life of existing servers, which has a big impact on reducing CO2. Melar Chen, product marketing manager at HashiCorp, argues that infrastructure as code (IaC) – which enables organisations to provision and manage infrastructure with configuration files, rather than through disparate workflows – offers IT departments a way to improve IT sustainability.
First, IaC makes it easier to collaboratively build, change and delete infrastructure in a safe, consistent and repeatable way. IT administrators can also create policies as code and automatic enforcement during the provisioning workflow, which ensures that best practices and security policies are not being violated. In addition, IaC offers auditing and a way to understand the implications of new or changed infrastructure before it is provisioned and applied.
When exploring areas to improve IT sustainability, there are often aspects of IT that organisations take for granted. But these can have a significant carbon footprint. One area often overlooked is bandwidth associated with web traffic and the impact of using multimedia-rich web content. According to Wholegrain Digital, which runs Website Carbon Calculator, the average web page produces 1.76 grams of CO2 per page view.
“For a smaller website with 10,000 monthly page views, that’s already 211kg of CO2 per year,” warns Tal Lev-Ami, co-founder and CTO of Cloudinary. However, as Lev-Ami points out, many e-commerce sites will have far more visitors, and organisations that rely on online sales naturally aim to increase, not reduce, web traffic.
However, he says the data transferred per web visitor can be reduced by addressing bandwidth. “Many companies are doing this already to reduce their costs and increase their web performance, but might not have analysed this from a CO2 perspective,” he says.
Quoting the American Council for an Energy-Efficient Economy, Lev-Ami says it takes 5.12kWh of electricity per gigabyte (GB) of transferred data. Given that the average US power plant produces 600g of carbon dioxide for every kilowatt-hour generated, according to US Department of Energy estimates, Lev-Ami says transferring 1GB of data produces 3kg of CO2.
Bandwidth is related to the number of visitors a website attracts and the size of the files comprising the websites that need to be downloaded each time a web page is rendered. By using image and video optimisation tools, it is possible to reduce the size of the data required by each web page.
By balancing images and video quality with file size, it is possible to reduce bandwidth. Lev-Ami says advanced image and video optimisation tools use AI to automate this process. He cites the case of one of Europe’s largest sportswear manufacturers, which reduced bandwidth consumption by 40% from 6.8TB (terabytes) per day to 4.05TB per day. Annualised, the company saved 618TB of bandwidth, which, according to Lev-Ami, equates to 1,890 tonnes of CO2 saved.
Read more about IT energy efficiency and sustainability
- Sustainability guides decision-making by incorporating economic, social and environmental impacts. Gartner looks at digital technology’s role.
- How can IT leaders separate fact from fiction when weighing up a tech supplier’s sustainability claims, and, crucially, what are the dangers or risks enterprises face if they do not do their due diligence?