Observability in the cloud

Riley Peronto at Chronosphere shares his insights on getting smarter about data and costs

 

Cloud-native architectures have transformed how organisations operate, bringing faster delivery, higher availability, and far more telemetry. As cloud-native architectures scale, observability data volumes increase by ten to a hundred times compared to traditional systems. In some cases, observability spend can outpace the cost of cloud infrastructure itself. 

 

For those planning or executing cloud-native adoption, the question is clear: how can the business maintain the visibility needed to deliver a great customer experience, while keeping costs in check? 

 

 

The cardinality conundrum: insight vs. cost

A big part of the answer lies in cardinality, which refers to the number of unique metric combinations an organisation’s systems emit. The total number of combinations with data that exist are cardinalities. Cloud-native architectures multiply those combinations rapidly, leading to high cardinality, which, in turn, exponentially increases the volume of telemetry data the organisation has to contend with.

 

When more dimensions are added to metrics and cardinality goes up, it quickly becomes not only costly but also difficult to navigate. Ultimately, the goal isn’t less data, it’s smarter data – captured at the right fidelity, for the right duration, and used with intent.

 

Engineers recognise that robust observability is a business necessity. In fact, strong observability is a key driver for innovation and the foundation for all business value. High-performing teams, therefore, treat observability like any other business-critical programme that is governed, measured, and continually optimised. A useful operating model is a four-step optimisation cycle: govern, analyse, refine, and operate.

 

 

Govern: enforcing controls from day one

Many organisations lack real-time visibility into which teams and applications consume the share of observability capacity. Cost allocation is often manual and delayed, leaving teams effectively blind to their consumption impact. A single team can overrun budgets before anyone notices.

 

Enforcing governance means creating logical partitions of capacity by team, service, or environment. From there, IT leaders can assign explicit quotas that will ensure that one group’s data spike can’t trample the rest.

 

More advanced observability practices go one step further, monitoring cost and usage attribution. This means that teams can monitor costs in the same way they monitor their production services: alerts fire as limits are approached. This enables them to avoid surprise bills, while there is still something that can be done about them.

 

 

Analyse: focus on value, not volume

Many observability tools offer blunt levers to reduce data. Organisations looking to cut costs can drop data on the floor or push it to a lower-cost tier. While these levers might be helpful in the short term, they also introduce risk: how can business leaders know their teams won’t need that data during a production incident?

 

That’s why sustainable cost control requires understanding how the various teams actually use the data. This means adopting platforms that demonstrate how data is utilised in dashboards, alerts, and queries. 

 

With these mechanisms in place, identifying quick wins becomes immediately obvious, and businesses can confidently reduce their data collection by eliminating unused metrics or duplicative metrics and logs that the team never references.

 

This will also lead to identifying data streams that can be optimised in more sophisticated ways, such as applying rollup rules to aggregate noisy, high-frequency series. These advanced rules preserve the visibility needed to deliver the ideal customer experience. 

 

 

Refine: precision is the key to success

To maintain visibility into production services, simply collecting less data is not a realistic option for businesses. Instead, businesses need to use the insight gained from mapping out their most important metrics to capture the right data in the right shape. 

 

That requires precise controls that enable teams to change fidelity and structure without sacrificing signal during an incident. Strong data controls are delivered centrally. The IT team ought to be able to drop, roll up, or transform metrics without needing to reinstrument or redeploy services, as well as preview changes so they can model the impact of a rule before enabling it. This capability allows for timely and safe responses during an incident and continuous optimisation in the meantime.

 

 

Operate: make optimisation continuous

Optimisation isn’t a one-off tidy-up; it’s a feedback loop. Once optimisation rules are in place, advanced observability practices will start measuring impact. This means tracking rule-level efficiency (such as percentage reduction in data points per second) and showing the downstream effect on observability costs. 

 

Ongoing visibility into the cost impact ensures that improvements stick. This will consist of real-time monitoring for cardinality spikes and cost anomalies, and maintaining dashboards that track the impact of optimisation over time, so regressions are caught early. 

 

Teams that bring governance, usage-aware analysis, and real-time control to observability achieve the outcomes that matter most to the board: lower, predictable costs and faster, more confident releases. Strong observability isn’t a luxury—it underpins business value. But value comes from clarity, not from collecting everything. By prioritising high-utility telemetry, shaping data at the source, and making optimisation continuous, organisations can keep cloud systems transparent and sustainable.

 

---

 

Riley Peronto is Director of product & solutions marketing at Chronosphere

 

Main image courtesy of iStockPhoto.com and Natali_Mis