Performance slowdowns of microservices or even application endpoints may stem from inefficient code or runtime issues like improper garbage collection and blocking threads. More severe problems like memory leaks can lead to application crashes. Since these problems are buried at the most granular level of the application, they are complex to diagnose and often require advanced expertise, lengthy reproductions, and disruptive application restarts. Moreover, such inefficiencies frequently lead to increase in cloud costs, as teams overprovision resources to compensate for poor performance.In this session, you will discover how Continuous Profiler—part of Datadog APM—provides deep visibility into the performance of production code and enables quick diagnosis of slow requests and runtime bottlenecks, as well as optimization of CPU usage and cloud spend. We will demonstrate how you can correlate APM spans with granular code- and thread-level behavior to identify issues like inefficient garbage collection and blocking threads. In addition, we will walk through a new guided workflow for detecting memory leaks, as well as ways to evaluate the performance implications of downsizing infrastructure to make more informed resource allocation decisions.