Observability on OCI: up to 40% lower telemetry cost

The challenge

Keeping a critical estate on Oracle Cloud Infrastructure (OCI) resilient and under control meant monitoring many moving parts at once: applications, databases, Kubernetes clusters and networks.

Without an integrated view, teams were slow to find the source of an incident, and diagnosis and resolution took longer. At the same time, ingesting and storing large volumes of telemetry that was not needed raised cost without necessarily improving diagnosis.

The need to shorten resolution times, contain rising telemetry cost and move toward proactive operations led to an end-to-end observability model.

The challenge, in one line: full visibility of critical systems, earlier detection, and a lower cost for the information required to operate.

The solution

We treated observability as three pillars: logs, metrics and traces. Each tool was chosen for the balance of diagnostic value, cost and performance.

1. Phase 1 · Logs

   We centralized audit data and events in OCI Log Analytics across OKE clusters, VCNs, VPNs, instances and DB Systems. Ingestion is filtered at the source, and cleanup policies avoid paying for information that is not needed.

2. Phase 2 · Metrics

   We complemented native OCI monitoring with VictoriaMetrics for time series from OKE and from Payara and Spring Boot services. OCI Functions inject tailored business metrics, and Database Management together with Ops Insights give deep visibility into the Oracle database.

3. Phase 3 · Traces

   OpenTelemetry carries application traces from OKE into OCI APM. Intelligent sampling, focused on critical transaction flows, follows distributed requests without saturating the network or raising processing cost.

4. Phase 4 · Unified visibility

   Early alerts and dashboards in the OCI console and in Grafana give engineering and SRE one place to correlate metrics, logs and database events.

The result

An estimated reduction of up to 40% in telemetry storage and ingestion cost, shorter diagnosis and resolution time (MTTR), and reported 100% observability coverage of the OKE clusters and OCI resources included in the project.

• Telemetry cost

  BeforeIngestion and storage of large volumes of unnecessary data, which increased operating cost.

  AfterAn estimated reduction of up to 40% in monthly ingestion and storage cost for telemetry that was not needed, through filtering and intelligent sampling.

• Incident resolution time (MTTR)

  BeforeSlower diagnosis, because information from applications, infrastructure and databases was hard to correlate.

  AfterShorter mean time to diagnose and resolve incidents in Spring Boot, Payara and Oracle DB, with an integrated view of logs, metrics and traces.

• Observability coverage

  BeforeLimited visibility and no integrated traceability across infrastructure components and applications.

  AfterReported 100% observability coverage of the OKE clusters and OCI resources included in the project, with centralized monitoring.

• Incident management

  BeforeA reactive model, with heavier reliance on manual investigation to find anomalies and determine their cause.

  AfterProactive management, with early alerts, centralized dashboards and event correlation that speed up the operational response.

Technology

• OCI Log Analytics
• OCI APM
• OpenTelemetry
• VictoriaMetrics
• OCI Functions
• Database Management
• Ops Insights
• Grafana
• OCI Dashboards
• Unified alarm engine

Testimonial