Profiling performance analysis – code-level performance bottlenecks and slow query management

Join the community

Scan with WeChat
Join the official community group

Try Guance

Start online with usage-based pricing and a true cloud service.

Get started

Choose a Guance edition

Free Business Enterprise

Code repositories

GitHub Gitee

Profiling performance analysis addresses what problems it solves

When Trace only shows slow service, Profile continues to locate code hotspots

APM Trace can tell the team which service segments are being called slowly, but it may not necessarily explain why. Profiling Performance Analysis samples data from the production environment, including function calls, CPU, memory, locks, IOs, and slow queries, further locating interface time, resource consumption, and user experience issues to specific code paths.

01Code hotspotsMethods to locate the highest CPU, memory, lock waiting, and IO consumption
02Flame diagram analysisUse the call stack view to clearly see function time and resource usage
03Trace associationJump from a slow request Span to the corresponding time period of Profile data

Interface is slow, but Trace only shows time consumption? Continue locating functions and resource consumption

Profiling can identify where function calls and resource consumption cause performance bottlenecks, such as excessive CPU usage, abnormal memory allocation, lock waiting, slow queries, or IO blocking. Teams can continue from "which interface is slow" to "which code segment consumes the most time and resources."

Book a demo View documentation

Use flame charts to clearly see CPU, memory, and IO hotspots, without guessing optimization

Guance Present Profiling stack data as flame diagrams and other methods, supporting method-level analysis of CPU, memory, lock waiting, and IO usage. R&D teams can analyze execution time by language, service, version, and time range to identify code paths that truly impact performance.

Book a demo View documentation

Jump from a slow request Span to the corresponding profile with one click, resulting in shorter root causes

Guance supports Profile data associated at the SPAN level. When reviewing slow requests or exception chains, you can obtain code hotspots, method call lists, and time usage proportions within the corresponding time window, helping the team determine whether the bottleneck stems from business logic, dependency calls, resource contention, or runtime issues.

Book a demo View documentation

Frequently asked questions

What is Profiling Performance Analysis?

Profiling helps teams identify code-level performance bottlenecks in production environments by sampling application runtime data on CPU, memory, lock waiting, IO, slow queries, and function calls.

What is the difference between Profile and APM Trace?

APM Trace mainly shows which services a request passes through and the time spent per segment; Profile further demonstrates code methods, call stacks, and resource consumption, helping explain why slow requests are slow.

What issues is profile analysis suitable for troubleshooting?

It is suitable for troubleshooting issues such as slow interfaces, high CPU usage, abnormal memory allocation, lock waiting, IO blocking, high function time consumption, and performance degradation in production environments.

Can Profile be linked to link tracing?

Yes, you can. Guance supports associating slow request Span with corresponding time period profile data, helping teams further target specific code hotspots for service time.

Explore more

Resources and further reading

Choose the next step from product documentation, related solutions, and technical guidance.

Related reading

Selected product practices, troubleshooting guidance, and technical solutions

Link tracing visualization tool: flame diagrams

This article demonstrates through real-world cases how flame graphs can be used to analyze service performance, identify problems, and take corresponding optimization measures.

Product capabilities

Guance best practices for link tracking analysis

This article uses RuoYi-Vue, a front-end and back-end separation application deployed on K8s, as a case study to explain how to perform link tracing analysis through Guance.

Best practices

Python Flask containerized application chain is observable

This article uses a containerized deployment example of the Python Flask application to explain how to perform link tracing through Guance to achieve performance monitoring and problem localization for applications.

Best practices

JAVA Application Implementation of Automatic APM Injection (Docker Edition)

This article explains how to inject APM probes into Java applications deployed by Docker through Guance.

Best practices