Making Software Greener: Hosting, Utilization and Measurement

At GOTO Aarhus 2023, Holly Cummins discusses software’s environmental footprint and practical ways to reduce wasted resources. She works on Red Hat’s Quarkus team and discloses that affiliation when discussing its measurements.

She organizes the problem around carbon-aware timing and location, hardware utilization and energy efficiency. Regional electricity mixes and workload requirements change, so a hosting decision needs current, location-specific data rather than a fixed ranking.

Her first suggestion is to choose hosting with electricity sources in mind and, where practical, schedule flexible workloads for lower-carbon periods. She notes that shifting work across locations adds orchestration and architectural challenges.

The second suggestion is to improve utilization and make systems able to turn off or scale down when they are not needed. Identifying idle resources requires visibility into usage and cost.

Cummins argues that optimization needs measurement. She discusses cost and carbon tools, while noting that cloud energy estimates, network use and emissions from manufacturing hardware can be difficult to capture.

Her third and fourth suggestions use runtime and spending as rough signals: faster execution or lower cost can sometimes indicate lower impact when other conditions are comparable. She explicitly calls them heuristics, not substitutes for a full footprint measurement; faster or cheaper is not always greener.

She presents her team’s Quarkus experiments comparing configurations and an alternative framework. In one workload, JVM throughput outweighed native compilation’s faster startup; the result is specific to the tested setup, not a universal ranking.

Her fifth suggestion is to use Quarkus, reflecting her role on that project. She ends with the broader idea of removing work that provides no benefit, such as obsolete CI jobs, while considering both operational and embodied impacts.