Green IT and observability: making the transition work

SECTION 1

Introduction

Organisations can draw on sustainability metrics and observability to advance their green IT initiatives. The premise is encouraging: it is often small changes that yield the greatest benefits, provided you measure and monitor your environmental impact methodically rather than by gut feeling.

Reducing the footprint of digital technology is not just a matter of goodwill. Without reliable data to objectify the situation, it is impossible to know where to act or to verify that efforts are paying off. This is precisely the role of metrics and observability: turning an intention into a measurable trajectory.

SECTION 2

Why green IT needs data, not just good intentions

### The hidden cost of digital technology

The environmental footprint of digital technology remains largely invisible to the teams that build and run systems: an oversized cluster, a nightly batch job with no real purpose, or an architecture that multiplies redundant network calls show up on no direct bill. Yet they translate into electricity consumption, greenhouse gas emissions, and ultimately financial cost, without triggering any classic alert signal.

### From measurement to action

It's this gap between real impact and operational visibility that green IT approaches aim to close. An organisation that doesn't measure can neither prioritise its efforts nor demonstrate their effectiveness to governance bodies or regulators. Conversely, as soon as a sustainability indicator is tied to a technical dashboard already consulted daily by operations teams, it becomes a concrete lever for action rather than an isolated CSR report produced once a year.

SECTION 3

The three metrics that structure sustainability measurement

### PUE and data center energy efficiency

Three metrics prove particularly enlightening. PUE, or power usage effectiveness, measures the energy efficiency of a data center by relating the total energy consumed by the facility to that actually used by IT equipment. A PUE close to 1 signals an infrastructure where cooling, power delivery, and ancillary losses are minimised; the best-performing data centers on the market in 2026 approach values around 1.1 to 1.2, versus older facilities that still exceed 1.6 to 1.8.

### Data throughput and resource utilisation rates

Data throughput assesses the volume processed over a given period, which must systematically be balanced against the associated energy consumption: processing more data isn't a problem in itself, as long as the energy-per-unit-of-useful-work ratio improves or stays stable. The third metric, resource utilisation rates, indicates how effectively software and hardware assets are used, a Kubernetes cluster sized for a peak it rarely hits, or cloud instances permanently provisioned at 15% load, are clear signals of waste. Together, these three indicators provide a reliable measurement baseline: they make it possible to establish a starting point, then compare the progress achieved over time.

SECTION 4

From monitoring to observability: understanding why, not just when

### A decisive distinction

Observability should not be confused with plain monitoring. Where monitoring identifies when a problem occurs, a consumption threshold breached, an alert on a load spike, observability explains why it happens, by correlating logs, metrics, and traces. This distinction is decisive for acting on causes rather than symptoms, and therefore for durably improving a system's energy efficiency rather than merely reacting to its occasional excesses.

### The techniques that make sustainability observable

Several techniques contribute to this real-time sustainability tracking: log analysis to spot redundant tasks or errors that trigger resource-costly retries, anomaly detection to identify a consumption drift before it becomes structural, and automated alerts to notify teams as soon as a service departs from its usual energy profile. Observability data, available continuously rather than as a one-off report, also helps justify green IT investments to decision-makers, with figures to back it up rather than convictions alone.

@cite:les-3-piliers-de-l-observabilite

SECTION 5

Building an observability stack for sustainability

### Instrumenting existing infrastructure and workloads

There's no need to start from a blank page: most organisations already have technical instrumentation, infrastructure metrics collection, log aggregation, distributed tracing, that can be enriched rather than replaced. The challenge is to add energy and carbon signals (consumption per node, estimated footprint per request, the power grid's carbon intensity factor by region and time of day) to the dashboards already used by SRE and platform teams.

### Linking carbon metrics to the technical metrics already tracked

The value of such a stack doesn't come from multiplying indicators, but from relating them: a latency spike correlated with an energy consumption spike, a performance regression that doubles the number of instances needed to sustain load, or a data-processing job whose carbon cost far outweighs its business benefit, all become visible and actionable as soon as they share the same observation plane as standard SRE metrics.

@cite:construire-une-pile-d-observabilite-proactive-avec-datadog

SECTION 6

The five obstacles to rolling out a green IT approach

### Human and organisational obstacles

Implementation runs into five recurring obstacles. The first is expertise gaps: digital sustainability calls for cross-disciplinary skills spanning engineering, finance, and CSR, rarely found together in a single team. The second is cultural resistance, which requires organisation-wide buy-in rather than just the goodwill of a pioneering team, without explicit management sponsorship, initiatives run out of steam within a few months.

### Technical, budgetary, and regulatory obstacles

On top of that come budget constraints in the face of the initial investment needed to instrument and modify existing systems, the difficulty of choosing observability tools genuinely dedicated to sustainability in a still-young and fragmented market, and compliance standards that vary by industry and geography, the European CSRD directive and its associated frameworks impose different levels of reporting granularity depending on an organisation's size and activity.

SECTION 7

Five steps towards a sustainable organisation

### Initial audit and a cross-disciplinary team

Five steps make it possible to progress despite these hurdles. The first is to conduct a system-wide sustainability audit covering infrastructure, applications, and development practices, to establish a quantified baseline rather than an impression. The second step is to assemble a cross-disciplinary team combining engineering, operations, procurement, and CSR, the only way to turn audit findings into cross-functional decisions.

### Metrics, SMART goals, and continuous tracking

Next comes choosing metrics and tools suited to the organisation's context, PUE, data throughput, utilisation rates, but also indicators specific to the industry. The fourth step sets SMART goals, specific, measurable, achievable, realistic, and time-bound, to avoid declarative ambitions with no deadline. The final step applies the changes and tracks progress continuously, relying on the observability stack built earlier rather than on an isolated annual audit. Cross-departmental collaboration and ongoing monitoring are the real drivers of long-term sustainability. At Adservio, we support organisations through this approach to embed green IT in their engineering and operations practices.

@cite:informatique-durable-construire-la-resilience

SECTION 8

Conclusion

Green IT can't be decreed: it has to be steered, with the same measurement and observability disciplines that have proven themselves in software reliability. Organisations that connect their sustainability metrics to their existing observability stack turn an environmental ambition into a verifiable, arbitrable, defensible trajectory, for their technical teams as much as for their external stakeholders.

FAQ

Frequently asked questions