Monitoring is the continuous process of collecting, tracking, and alerting on predefined metrics that reflect the health, uptime, and performance of a system running in production. It is a foundational practice in DevOps and site reliability engineering, giving teams a structured way to know whether their applications and infrastructure are behaving as expected at any given moment.
At its core, monitoring works by measuring specific, known indicators over time. These typically include server CPU and memory usage, response times, error rates, request throughput, and uptime. When one of these metrics crosses a defined threshold, the monitoring system triggers an alert, notifying the relevant team so they can investigate and respond before end users are significantly affected. This reactive-but-informed approach is what makes monitoring essential to maintaining service reliability.
What Monitoring Tracks
Production monitoring generally spans several layers of a system. Infrastructure monitoring covers the underlying servers, containers, or cloud resources. Application monitoring focuses on the behavior of the software itself, tracking things like slow database queries, failed API calls, or elevated HTTP error rates. Uptime monitoring, also called availability monitoring, checks at regular intervals whether a service is reachable and responding correctly. Together, these layers give engineering teams a comprehensive view of system status.
Monitoring vs. Observability
Monitoring is frequently discussed alongside observability, but the two concepts are distinct. Monitoring answers the question "is something wrong?" by checking predefined metrics against known thresholds. Observability goes further, asking "why is something wrong?" by enabling engineers to explore the internal state of a system through logs, traces, and metrics in combination. In practical terms, monitoring tells you that error rates have spiked; observability gives you the tools to trace exactly which service, function, or dependency caused it.
This distinction matters because modern distributed systems, built on microservices and cloud infrastructure, can fail in ways that predefined metrics alone do not anticipate. Monitoring remains indispensable for catching known failure modes quickly, while observability handles the unknown unknowns. The two practices are complementary rather than competing.
Monitoring and APM
Application Performance Monitoring, commonly abbreviated as APM, is a specialized form of monitoring focused specifically on the performance and availability of software applications. APM tools typically combine monitoring with deeper diagnostic capabilities, making them a bridge between traditional monitoring and full observability. Platforms such as Datadog, New Relic, and Dynatrace are widely used examples in this space.
For development and operations teams, establishing solid monitoring is usually the first step toward production reliability. Without it, there is no reliable signal that a deployment has degraded performance or that a service has gone down. With it, teams can set meaningful uptime targets, respond to incidents faster, and build confidence that their systems are serving users as intended.