/Glossary · DevOps
What is Observability?
Observability is the ability to understand a system's internal state from the data it emits, mainly metrics, logs and traces. It lets you answer new questions about a problem without having to deploy additional code.
How observability works
Metrics show aggregate trends, such as latency or errors per minute. Logs record detailed events with context. Distributed traces follow a request through every service it touches. Correlating all three signals with shared identifiers lets you go from symptom to cause in minutes.
On AWS, the core is Amazon CloudWatch for metrics, logs and alarms, AWS X-Ray and CloudWatch Application Signals for traces and service views, and AWS Distro for OpenTelemetry for standard instrumentation. For Kubernetes, add Amazon Managed Service for Prometheus and Amazon Managed Grafana.
When to invest in observability
When an incident takes hours to diagnose, when the first sign of a problem comes from a customer, or when a microservices architecture makes it impossible to follow a request. The more distributed the system, the more you need observability to operate it with confidence.
It is also worth doing before you scale or migrate: without a performance baseline you can't tell whether a change made things better or worse. And it is a practical prerequisite for adopting SRE, because SLOs need reliable indicators measured from the user's point of view.
Best practices and common mistakes
Instrument with OpenTelemetry to avoid vendor lock-in, use structured JSON logs and propagate a trace ID across every call. Set retention by data type and use lower-cost log classes for what you rarely query. Build per-service dashboards focused on user impact.
Common mistakes: logging everything with no structure and overpaying for ingestion, creating hundreds of alarms nobody responds to, and not correlating signals with each other. Another is instrumenting only the infrastructure and not the business flows, such as a payment or a sign-up.
/FAQ
Questions about Observability
Another question? Write to hello@nubent.com and an engineer will answer.
What is the difference between monitoring and observability?
Monitoring answers questions you already know to ask, like whether CPU crosses a threshold. Observability lets you investigate problems you didn't anticipate by exploring correlated metrics, logs and traces. Monitoring is part of observability, not its opposite.
What is OpenTelemetry?
It is an open CNCF standard for generating and exporting metrics, logs and traces. AWS offers AWS Distro for OpenTelemetry, which lets you send that data to CloudWatch, X-Ray or other tools without changing the instrumentation in your code.
How do you control observability costs on AWS?
Set retention per log group, filter out low-value events before sending them, sample traces and consider the CloudWatch Logs Infrequent Access class. Also review high-cardinality custom metrics, which often drive unexpected costs.
How we help with Observability
Observability and SRE
CloudWatch, OpenTelemetry and Grafana, SLOs with error budgets, actionable alerts, incident management and postmortems to run AWS on data.
Learn moreIncidents and availability
AWS incident response and high availability: stabilization, root cause analysis, Multi-AZ architecture and disaster recovery with defined RTO and RPO.
Learn moreServerless on AWS
Lambda, API Gateway, Step Functions, EventBridge, DynamoDB and SQS: event-driven architectures that scale on their own, tolerate failures and bill only for real usage.
Learn morePerformance and scalability
AWS performance and scalability: load testing, bottleneck analysis and tuning of the database, caching and auto scaling, measured before and after.
Learn moreNext step
What should we build next on AWS?
30 minutes with an engineer. Concrete next steps, whether or not you work with us.