Alibaba‘s ARTHAS

Alibaba's ARTHAS is a powerful Java diagnostic tool designed for real-time troubleshooting and performance analysis of Java applications. It supports deep-dive analysis of JVM issues and provides a variety of commands to analyze method performance. Here's how you can use Arthas to analyze method performance:

---

Step-by-Step Guide to Analyze Method Performance Using Arthas

1. Install Arthas

• Download the Arthas package:

  curl -O https://arthas.aliyun.com/arthas-boot.jar

• Start Arthas:

  java -jar arthas-boot.jar

• Follow the prompts to attach Arthas to your running Java application.

---

2. Attach to the Target Application

When you start Arthas, it lists all running Java processes. Select the target process by typing the corresponding number.

---

3. Use Key Arthas Commands

A. Method Execution Time Analysis (trace)

The trace command tracks the execution time of a method, including calls to other methods.

Example:

trace com.example.MyService myMethod

• Outputs the execution time of myMethod and its internal calls.
• Useful for identifying slow methods or bottlenecks.

---

B. Execution Statistics (monitor)

The monitor command provides statistical data like invocation count, success rate, and average response time.

Example:

monitor com.example.MyService myMethod

• Displays metrics such as:
  • Total invocation count.
  • Maximum, minimum, and average execution time.
  • Error rate.

---

C. Profiling Method Performance (profiler)

The profiler command generates a flame graph to visualize method execution.

Steps:

1. Start profiling:

   profiler start

2. Perform the action you want to analyze.
3. Stop profiling and save the report:

   profiler stop

• View the flame graph to identify hotspots in your application.

---

D. Observe Input and Output Parameters (watch)

The watch command captures and displays the input, output, and exceptions of a method.

Example:

watch com.example.MyService myMethod '{params, returnObj, throwExp}' -x 2

• Displays:
  • params: Input parameters.
  • returnObj: Output of the method.
  • throwExp: Any exceptions thrown.

---

E. Monitor Method Execution in Real Time (tt)

The tt command (Time Tunnel) records method calls and allows you to replay them for debugging.

Steps:

1. Record method calls:

   tt -t com.example.MyService myMethod

2. Replay recorded calls:

   tt -i <index>

• Analyze the method's behavior in detail, including parameters and return values.

---

F. Method Invocation Stack Analysis (stack)

The stack command shows the stack trace of a method invocation.

Example:

stack com.example.MyService myMethod

• Identifies where a method is being called from and the sequence of calls leading to it.

---

G. JVM and System Metrics (dashboard)

Use the dashboard command to monitor JVM performance metrics like CPU usage, memory consumption, and thread count.

Example:

dashboard

• Provides an overview of the application's runtime environment.

---

4. Analyze Data

• Use trace, monitor, or watch to pinpoint slow or inefficient methods.
• Use profiler to identify hotspots visually.
• Analyze stack traces with stack to understand method dependencies and call hierarchies.

---

5. Optimize and Re-Test

• Refactor inefficient methods based on the insights.
• Use Arthas again to verify the performance improvements.

---

Best Practices

1. Start with dashboard to get a high-level overview.
2. Use trace and monitor for pinpointing bottlenecks.
3. Visualize hotspots with profiler.
4. Avoid attaching Arthas to a production environment without proper safeguards (e.g., load impact considerations).
5. Combine Arthas insights with other tools like JProfiler or VisualVM for comprehensive analysis.

---

Arthas simplifies real-time performance debugging and is ideal for production environments.