Application performance monitoring (APM) has historically involved a lot of hunting and educated guesswork. If performance deteriorated, monitoring teams would investigate factors like CPU, RAM and storage availability in hopes of identifying the culprit. This often led to dead ends because the root of performance problems was somewhere else. Disparate data points were often displayed on multiple screens, requiring operators to correlate information manually. And problems that weren't easily identified by infrastructure monitoring were nearly impossible to detect.

APM is being redefined by innovations in performance monitoring and a new perspective that places user experience at the center of the equation

Now, APM is being redefined by innovations in performance monitoring and a new perspective that places user experience at the center of the equation. Instead of requiring operators to constantly query the system about its status, modern observability solutions continually display the state of the system as part of normal operations. Visualizations enable operators to see problems quickly, in some cases even before they manifest themselves in a degraded user experience. In short, traditional APM is reactive while modern approaches are proactive and predictive.

There is a clear demand for APM's insights. According to New Relic's 2023 Observability Forecast, more than half (53%) of survey respondents had deployed APM, a 17% increase year-over-year. Nine in 10 (89%) expected to deploy APM by 2026. The monitoring is working. More than two-thirds (69%) of those who currently deploy APM said their organization's MTTR improved since adopting observability, including 35% who said it improved by 25% or more.

Observability solutions now peer into the deepest recesses of applications, uncovering every factor that may affect performance. These include such new cloud-native variables as the health of software containers, tool- and language-specific characteristics, connectors to external data sources, custom integrations, and application program interfaces.

A Complete Picture

The latest generation of APM tools can trace an intricate web of interconnected services to unmask the threads of communication that tie them together. Auto-discovery identifies new applications and code deployments and automatically incorporates them into the fabric of services being monitored. Machine learning observes the factors that affect the performance of individual applications over time and learns to look for changes that presage a slowdown or outage.

A critical feature of today's solutions is an integrated dashboard that enables operators to view such useful troubleshooting aids as distributed traces — which track interactions within complex systems — alongside APM telemetry. They look for significant incidents that influence performance and continually aggregate log information into clusters that allow patterns to be observed without the need for administrators to search or scan through thousands of log entries. Coordinated timestamps correlate changes in performance with possible causal factors and enable operators to drill down on anomalies for problem detection and resolution.

The result is a view of application performance from both above and below. At the center of the operator view are the metrics that are most critical to the user experience, such as response and load times. Alongside that are summaries of alerts, deployments, service levels and vulnerabilities, which are the most critical factors in diagnosing performance problems.

If a spike in response times is detected, operators can scroll down to look at elements of infrastructure, dependencies, databases, containers and other services. By viewing distributed traces alongside APM telemetry, they can quickly identify the root cause of service issues and navigate to the relevant trace to further investigate the problem. They can even drill into the application code to spot problematic changes and see when they were introduced.

This doesn't mean traditional metrics are no longer needed. They are still a great way to identify common infrastructure problems such as bad memory or a corrupt database table. The difference with redefined APM is that the customer experience is at the center and all the factors that affect it are tied to that crucial metric. The latest solutions also enable rich integrations with third-party solutions as well as connections to the vast collection of APIs, software development kits and tools available in the OpenTelemetry observability framework.

Organizations don't have to worry about their APM solutions becoming obsolete but can focus on what really matters: Delighting users.