Green ICT: Optimizing Systems for Efficiency

Green ICT isn’t about spending more on maintenance or sacrificing performance. It’s about developing systems and infrastructure thoughtfully.

Imagine a bustling online store catering to a global audience, especially during peak events like Black Friday. This store operates on Azure using Kubernetes via AKS. Initially lacking autoscaling capabilities, the store responded to slowdowns by adding numerous virtual machines without a clear understanding of CPU or memory usage. This led to reliance on costly infrastructure, with 14 units for adequate computational power, even during quieter periods. It was evident that an optimization strategy emphasizing efficiency and sustainability was needed.

To reduce expenses and cut CO2 emissions, the first step was understanding the application’s functionality. Fortunately, the application’s ‘stateless’ nature ensured continuity in shopping cart data across virtual machines. A deep dive into resource usage unveiled inefficiencies. Peak times ideally see around 60% CPU and memory usage, but stress testing revealed usage rates far below these thresholds, even with larger instances. It can be reasonably stated that the application's behavior had potential for enhancements.

Understanding the application’s behavior prompted the introduction of auto-scaling, allowing multiple containers on a single virtual machine—perfectly aligned with Kubernetes’ stateless nature and flexibility in varied environments. A Kubernetes test cluster facilitated fine-tuning without reliability concerns. This paved the way for implementing the production AKS cluster, optimizing CPU and memory usage while validating autoscaling during traffic surges. These optimizations resulted in improved testing process, quicker feedback cycles, minimized production issues, and overall team satisfaction. Optimization techniques, including enhanced caching and reduced reliance on third-party APIs, significantly enhanced system performance.

As a summary, all these changes reduced costs but also CO2 emissions. The transformation from 14 costly virtual machines (€7000/month) to 10-16 more efficient ones (€1300/month) underscores a crucial point: ‘Green ICT’ isn’t synonymous with ‘Expensive ICT.’ Instead, it emphasizes better resource utilization and cost consciousness.

Read more about this topic in Teemu´s blog