Metrics are numerical measurements collected at regular intervals across your applications, servers, networks and cloud platforms. They give engineering and operations teams a fast, lightweight way to track system health, spot performance trends and trigger reliable alerts.
Every system generates continuous signals. CPU load climbs, memory fills up, response times drift or error rates spike during peak traffic.
Metrics capture these shifts as compact numbers over time. Because they are lightweight to store and query, metrics power your primary dashboards and automated alerts. Yet a rising number only tells you that something changed. Pinpointing why it changed still requires connecting metrics with traces and logs.
A metric data point captures four core details at each measurement interval:
The specific measurement being tracked, such as HTTP request duration or CPU usage.
The recorded number showing system state at that exact moment in time.
The precise date and time when the measurement was captured and stored.
Metadata attributes like service name, host, region or status code that let you slice and filter results.
http_request_duration_ms{service="payment_api", region="eu_north", status="500"} 482.5 @ 2026-10-22 14:32:00
This shows that payment API latency jumped alongside server errors in one region. To uncover the exact failing request or database error behind that spike, you pivot from this metric to traces and logs.
How hardware and hosts perform
Track CPU load, memory, disk space and network traffic across servers, containers and cloud hosts.
How software responds
Measure request latency, throughput, error rates and queue depth across microservices and APIs.
How data layers hold up
Monitor query latency, connection pool usage, cache hit ratios and replication lag under load.
How execution engines behave
Track garbage collection pauses, thread counts, container restarts and Kubernetes pod health.
How outcomes match expectations
Measure completed checkouts, sign up rates and payment volumes alongside technical performance.
As environments scale across containers and microservices, metric volume multiplies quickly. Without clear standards, teams run into familiar obstacles:
Unbounded labels like user IDs or full URLs create millions of unique series and inflate costs.
Static thresholds trigger noisy alerts on brief spikes that recover on their own.
Services report latency in varied units and names, breaking shared dashboards.
Long collection intervals smooth out short bursts of latency or packet loss.
Teams switch between separate infrastructure and application views without shared service labels.
Measure latency, traffic, errors and saturation first to reflect real user experience.
Keep tags bounded to meaningful dimensions like service, environment and region.
Use shared semantic conventions so every service reports metrics in consistent units and names.
Tie alerts to sustained error budgets and user impact rather than brief resource blips.
Carry shared service and trace tags so engineers move straight from an alert to root cause.
Metrics are one of the three core signals of observability:
With OTel, applications generate standardised metrics through language SDKs or host receivers and route them through the OpenTelemetry Collector. Teams can aggregate, filter and export clean time series data to platforms like Elastic without vendor lock.
Observata helps your team design clean metric pipelines, control cardinality and align dashboards across Elastic and OpenTelemetry. We turn raw numbers into reliable operational insight.
Explore our approachOpen standards for collecting metrics, logs and traces. Discover how OpenTelemetry connects telemetry without vendor lock.
Read more about What is OpenTelemetry?Timestamped event records that detail application activity, errors and system state. Learn how structured logs turn raw data into reliable answers.
Read more about What are logs?Distributed request paths that follow transactions across services and dependencies. See how traces pinpoint where delays and failures happen.
Read more about What are traces?