Sustainability as an Engineering Discipline in Cloud Native Systems
Sustainability Is an Architecture Problem, Not a Slogan
Most engineering teams I know can tell you their p99 latency to the millisecond. Very few can tell you what a single microservice costs the planet to run. We measure what we are rewarded for, and for twenty years the industry has rewarded speed and availability, with efficiency a distant third.
The signal
The CNCF community recently wrapped up Cloud Native Sustainability Month. The sessions covered carbon observability with OpenTelemetry, energy metrics, Kubernetes resource allocation, eBPF, GPU efficiency, and how data center efficiency is calculated. The organizers’ main takeaway was that sustainability starts with questioning how we use the resources we already have, and that you don’t need to be an expert to join the conversation.
Over-provisioning is the original sin
Here is what I find interesting. Notice what the topics have in common: none of them is about building something new. They are all about visibility and discipline in how existing infrastructure is consumed.
Kubernetes made provisioning easy, and ease breeds waste. Teams pad CPU requests and memory limits “just to be safe”. Nobody gets paged for an idle node, but everybody gets paged for an outage. The incentives are asymmetric, so the rational local decision is to over-allocate, and the aggregate result is enormous stranded capacity burning electricity.
That is why I think the tooling mentioned here matters. eBPF gives kernel-level telemetry with low overhead. Energy-attribution projects like Kepler estimate power consumption per pod. OpenTelemetry offers a path to put carbon signals next to latency and error rates. Taken together, they turn sustainability from a yearly ESG report into an engineering signal that sits in the same dashboard as everything else. Once a metric is on the dashboard, it enters sprint conversations.
The trade-offs nobody likes to discuss
I’d caution CTOs against treating this as a free lunch. Right-sizing workloads competes with resilience headroom. Aggressive bin-packing can raise noisy-neighbour risk. Energy measurement is often modelled rather than metered, so the numbers are estimates, and acting on a poor estimate can create false confidence.
GPUs sharpen the dilemma. With generative AI adoption, the “just add another GPU” reflex is becoming the new over-provisioning. I have seen organizations rush to reserve accelerators for workloads that a smaller model, quantization, or batch scheduling would handle at a fraction of the energy. The architectural question should come before the procurement question: do we need this much compute, or have we simply never measured?
My view is that efficiency and cost converge. Every wasted core is both a carbon cost and a cloud bill. Sustainability becomes easiest to sell internally when you frame it as FinOps with a conscience.
A lesson for Bharat’s builders
This applies directly to us. Many startups and MSMEs in India, and certainly in the Northeast, run on tight budgets and uneven power and connectivity. Frugality is already a survival skill here. If we pair it with observability, we can build lean systems by design rather than by compulsion, and that is a real competitive edge as global buyers start asking suppliers for carbon disclosures. Students should learn resource-aware engineering as a core skill, not an elective.
Takeaways
- Measure before you optimize. Put energy and resource-utilization metrics beside your SLOs.
- Challenge defaults. Audit resource requests, idle environments, and GPU reservations quarterly.
- Treat estimates honestly. Modelled energy data is directional, so use it to find outliers, not to claim precision.
- Start small. A lightning talk or a single dashboard in your community or team is a legitimate first step.
- Link it to cost. Efficiency wins survive budget reviews when they save money too.
Closing thought
The most sustainable system is rarely the cleverest one. It is usually the one whose builders were curious enough to ask, “Do we actually need this?”
About the Author: Sanjeev Sarma is the Founder Director and Chief Software Architect at Webx Technologies. With a core focus on Generative AI integration, Cloud-Native Scalability, and Enterprise Software Architecture, he has spent over two decades driving digital transformation across Northeast India and beyond. Beyond his corporate leadership, Sanjeev is deeply invested in shaping the future of the IT industry. He serves as an Industry Expert on the Board of Studies for Assam Don Bosco University’s School of Technology, advises state technology committees, and actively mentors emerging tech startups at STPI. He brings a unique, dual perspective of high-level enterprise execution and future-ready academic curriculum development.