Cloud computing provides an inherent sustainability advantage over on-premises infrastructure: shared resources achieve higher utilisation rates, and major cloud providers invest heavily in renewable energy. But running on the cloud does not automatically make your architecture sustainable. Poorly designed cloud applications can waste as much energy as poorly managed data centres.
The first principle of sustainable cloud architecture is right-sizing. Most cloud workloads are over-provisioned, running on instance types larger than needed because the team provisioned for peak load and never revisited the decision. Auto-scaling addresses this by matching capacity to demand, but only if the scaling thresholds and instance sizes are correctly configured. We review scaling configurations quarterly and right-size based on the 95th percentile of actual usage.
The greenest compute is the compute you do not run. Design for efficiency, not excess capacity.
Serverless architectures are inherently more energy-efficient than always-on infrastructure because they consume compute resources only when processing requests. A Lambda function that runs for 200 milliseconds per invocation uses energy only during those 200 milliseconds. An EC2 instance serving the same function consumes energy 24 hours a day, even during periods of zero traffic. For workloads with variable or unpredictable traffic patterns, serverless is the most energy-efficient option.
Architectural Patterns for Sustainability
Data management is often the largest energy consumer in cloud applications. Storing data costs energy for disk operations and replication. Querying data costs energy for CPU and I/O. We apply data lifecycle management rigorously: archiving old data to cold storage, compressing data in transit and at rest, and designing schemas that minimise the data scanned by queries. A query that reads 1GB of data to return a 1KB result is wasting energy on a massive scale.
Region selection also matters. AWS publishes sustainability data for each region, and the carbon intensity varies significantly. The eu-west-1 region in Ireland runs primarily on renewable energy, while some other regions rely more heavily on fossil fuels. Where data residency requirements allow, deploying in regions with lower carbon intensity reduces the environmental impact of your workload.
- Right-size all cloud resources based on actual utilisation data
- Use serverless architectures for variable or unpredictable workloads
- Implement data lifecycle management to reduce storage and query costs
- Choose cloud regions with lower carbon intensity where data residency allows
- Include sustainability metrics in architecture review processes
- Use auto-scaling with properly configured thresholds to match capacity to demand
We include sustainability metrics in our architecture review process. For every new system, we evaluate the expected energy profile and identify opportunities to reduce consumption through architectural choices. This is not about perfection. It is about making informed decisions that consider environmental impact alongside performance, cost, and reliability.
