Green IT & Sustainable Software

The amount of CO₂ equivalent emitted per unit of electricity generated (gCO₂eq/kWh) — varies by grid, region, and time of day depending on the mix of energy sources (coal, gas, wind, solar, nuclear) in use.

"Our carbon-aware scheduler checks the grid carbon intensity API and defers batch ML training jobs to night hours when our region's intensity drops to 50 gCO₂eq/kWh versus 400 during peak demand."

A data centre efficiency metric: total facility energy divided by IT equipment energy. A PUE of 1.0 is perfect (all energy goes to compute); typical modern data centres target PUE below 1.4.

"Our on-premises data centre had a PUE of 2.1 — for every watt of compute, we spent another 1.1 watts on cooling and power distribution. Migrating to a hyperscaler with PUE 1.15 reduced our total energy consumption by 36%."

The practice of shifting or shaping workloads in time or geography to use electricity when and where it has the lowest carbon intensity — running compute when the grid is cleanest.

"We implemented carbon-aware computing for our video transcoding pipeline: jobs are queued and dispatched to whichever cloud region has the lowest carbon intensity at that moment, reducing emissions by 40% with no impact on SLA."

A GHG Protocol framework classifying emissions: Scope 1 = direct emissions (on-site generators, company vehicles), Scope 2 = purchased electricity, Scope 3 = indirect value chain emissions (employee travel, supply chain, cloud usage by customers).

"Our Scope 2 emissions fell 80% when we switched to a cloud provider running on 100% renewable energy. But our Scope 3 analysis revealed that employee laptop manufacturing and disposal accounted for more carbon than our entire cloud bill."

The principle that a computing device should consume power in proportion to the work it is doing — ideally zero power at zero utilisation and full power at full utilisation — minimising idle energy waste.

"Our old servers consumed 60% of full power even at 5% CPU utilisation — they weren't energy-proportional. Moving to modern ARM-based instances that idle near 10% power cut our per-request energy cost by 55%."

The practice of applying operational excellence principles to sustainability goals — measuring, optimising, and continuously improving the environmental footprint of IT systems alongside cost and performance.

"Our GreenOps dashboard tracks energy per API call, carbon intensity of our workloads, and idle resource waste alongside the usual cost and reliability metrics — sustainability is now a first-class operational concern."

Electronic waste — discarded electrical or electronic equipment including servers, laptops, phones, and networking hardware. The fastest-growing waste stream globally, containing both hazardous materials and recoverable precious metals.

"Our hardware refresh policy now requires certified ITAD vendor processing for all e-waste — we get a destruction certificate and a report on materials recovered. Previously decommissioned servers were going to landfill."

The process of safely disposing of or repurposing obsolete IT equipment — including data destruction, refurbishment, resale, and certified recycling — to comply with data security requirements and reduce e-waste.

"Our ITAD process extends server life by three years through refurbishment before recycling: a 5-year-old server refurbished and reused avoids 85% of the carbon cost of manufacturing a replacement."

An economic model that eliminates waste by keeping products and materials in use for as long as possible through reuse, repair, remanufacturing, and recycling — contrasted with the linear 'take-make-dispose' model.

"Applying circular economy principles to our IT estate: we standardised on a single laptop model to make repairs feasible, set up an internal refurbishment workshop, and extended average device life from 3 years to 5 — cutting hardware-related carbon by 40%."

The carbon emissions produced during the manufacture, transport, and disposal of hardware — as distinct from operational carbon (energy used while running). For many devices, embodied carbon exceeds lifetime operational carbon.

"Our lifecycle analysis showed that embodied carbon from manufacturing a new server was 3× the operational carbon over its 5-year life — this shifted our strategy from frequent hardware refresh to maximising hardware utilisation and lifespan."

A market-based instrument representing one megawatt-hour of electricity generated from a renewable source — purchased by organisations to claim renewable electricity usage even if the grid they draw from is not purely renewable.

"Our cloud provider claims 100% renewable energy through RECs — we audited whether they use time-matched RECs or annual averages. Annual RECs allow claiming renewable power while drawing from coal at night; time-matched RECs are a stronger standard."

Coding and architecture practices that reduce the energy consumption of software — including reducing unnecessary computation, minimising data transfer, choosing efficient algorithms, and designing for hardware longevity.

"Applying green software patterns to our mobile app reduced data transfer by 60%: we compressed payloads, cached aggressively, and eliminated polling in favour of push. Battery life for users on our platform improved measurably — a win for sustainability and engagement."