How does Jamm compare to Java VisualVM for memory measurement?

Jamm provides object-level memory measurement including JVM overhead and deep object graphs, whereas VisualVM focuses on heap dumps and high-level monitoring. Jamm is better for detailed optimization, but VisualVM is more user-friendly for general profiling.

How to measure memory usage of ByteBuffer slabs in Jamm?

Use MemoryMeter.builder() and call measureDeep with ByteBufferMode set to SLAB_ALLOCATION_SLICE or SLAB_ALLOCATION_NO_SLICE, depending on how slabs are created (e.g., with slice() or duplicate()), to accurately omit shared buffer overhead.

Does Jamm work with Java 17 records and new features?

Yes, Jamm 0.4.0 supports Java 17, but for records and hidden classes like lambdas, the Unsafe strategy may delegate to Specification, and using the Java agent is recommended for best accuracy due to reflection limitations.

Can I skip measuring certain fields without modifying the source code in Jamm?

Yes, you can use custom filters through the MemoryMeter builder or apply the @Unmetered annotation at runtime if allowed, but annotation requires class access, so filters offer more flexibility for external classes.

What's the performance overhead of using Jamm in production?

Jamm can have significant overhead, especially with deep object graph traversal using reflection or Unsafe, and it tracks visited objects, adding memory cost. It's designed for profiling, not continuous monitoring, so use it cautiously in production.

Jamm or Apache Commons SizeOf for memory measurement?

Jamm is more accurate and up-to-date, with support for modern Java versions and deep graph analysis, while Commons SizeOf is simpler and lighter but may be less precise for complex object structures or JVM overhead.

How to fix Jamm reflection errors in modular Java applications?

Add JVM arguments like --add-exports or --add-opens to allow reflective access. For example, --add-opens java.base/java.lang=ALL-UNNAMED might be needed for certain classes, but this can complicate deployment and security.

jamm — Java Object Memory Measurement

What is jamm?

Jamm is a Java library and agent that measures the actual memory usage of Java objects, including JVM overhead. It provides the `MemoryMeter` tool to help developers understand the true memory footprint of their objects, which is essential for performance tuning and avoiding memory leaks in Java applications.

Target Audience

Java developers and performance engineers who need to analyze and optimize memory consumption in their applications, particularly those working with complex object graphs or large-scale systems.

Value Proposition

Jamm offers accurate memory measurement across multiple Java versions (8, 11, 17) and JVM configurations, with support for various measurement strategies and flexible filtering options, making it a reliable tool for deep memory analysis.

Java Agent for Memory Measurements

Use Cases

Best For

Profiling memory usage of complex object graphs in Java applications
Optimizing memory consumption in performance-critical systems
Debugging memory leaks by understanding object footprints
Comparing memory overhead of different data structures
Benchmarking memory impact of code changes across Java versions
Analyzing ByteBuffer allocations in slab-based memory management

Not Ideal For

Applications deployed on non-HotSpot JVMs like OpenJ9, due to untested compatibility and assumptions about JVM internals
Projects where JVM startup cannot be modified to include the -javaagent flag, limiting access to the most accurate Instrumentation strategy
Environments with strict Java module system (JPMS) enforcement that block reflective access without additional JVM arguments like add-exports
Quick prototyping or development phases where only high-level memory estimates are needed, as Jamm's deep analysis adds overhead

Pros & Cons

Pros

Accurate Measurement Strategies

Supports Instrumentation, Unsafe, and Specification strategies, with Instrumentation being safest when using the Java agent for precise object size estimation across Java 8, 11, and 17.

Deep Object Graph Analysis

Can recursively measure entire object graphs with measureDeep, including options to filter classes or ignore references, essential for profiling complex memory footprints.

Flexible Filtering Options

Allows skipping measurement of specific classes or fields using the @Unmetered annotation or custom filters, providing fine-grained control over memory analysis.

ByteBuffer Specialization

Offers specialized modes (NORMAL, SLAB_ALLOCATION_NO_SLICE, SLAB_ALLOCATION_SLICE) for accurate ByteBuffer measurements in slab-based allocations, critical for buffer management scenarios.

Cons

Breaking API Changes

Version 0.4.0 introduced significant breaking changes, such as removing constructors and static methods in favor of a Builder pattern, requiring code updates and migration efforts.

HotSpot JVM Dependency

Assumes and only tested with HotSpot JVM; explicitly not compatible with other JVMs like OpenJ9, limiting its use in diverse or legacy environments.

Java Module System Complications

Under JPMS, reflection may be blocked, requiring add-exports or add-opens JVM arguments for accurate crawling, which can be cumbersome and error-prone in production setups.

Inaccuracies Without Agent

Without the -javaagent, strategies like Unsafe may fail for records or hidden classes (e.g., lambdas), and Specification has known issues with certain JVM arguments, reducing reliability.

Frequently Asked Questions

What is jamm?

Target Audience

Java developers and performance engineers who need to analyze and optimize memory consumption in their applications, particularly those working with complex object graphs or large-scale systems.

Value Proposition

Use Cases

Best For

Profiling memory usage of complex object graphs in Java applications
Optimizing memory consumption in performance-critical systems
Debugging memory leaks by understanding object footprints
Comparing memory overhead of different data structures
Benchmarking memory impact of code changes across Java versions
Analyzing ByteBuffer allocations in slab-based memory management

Not Ideal For

Applications deployed on non-HotSpot JVMs like OpenJ9, due to untested compatibility and assumptions about JVM internals
Projects where JVM startup cannot be modified to include the -javaagent flag, limiting access to the most accurate Instrumentation strategy
Environments with strict Java module system (JPMS) enforcement that block reflective access without additional JVM arguments like add-exports
Quick prototyping or development phases where only high-level memory estimates are needed, as Jamm's deep analysis adds overhead

Pros & Cons

Pros

Accurate Measurement Strategies

Supports Instrumentation, Unsafe, and Specification strategies, with Instrumentation being safest when using the Java agent for precise object size estimation across Java 8, 11, and 17.

Deep Object Graph Analysis

Can recursively measure entire object graphs with measureDeep, including options to filter classes or ignore references, essential for profiling complex memory footprints.

Flexible Filtering Options

Allows skipping measurement of specific classes or fields using the @Unmetered annotation or custom filters, providing fine-grained control over memory analysis.

ByteBuffer Specialization

Offers specialized modes (NORMAL, SLAB_ALLOCATION_NO_SLICE, SLAB_ALLOCATION_SLICE) for accurate ByteBuffer measurements in slab-based allocations, critical for buffer management scenarios.

Cons

Breaking API Changes

Version 0.4.0 introduced significant breaking changes, such as removing constructors and static methods in favor of a Builder pattern, requiring code updates and migration efforts.

HotSpot JVM Dependency

Assumes and only tested with HotSpot JVM; explicitly not compatible with other JVMs like OpenJ9, limiting its use in diverse or legacy environments.

Java Module System Complications

Under JPMS, reflection may be blocked, requiring add-exports or add-opens JVM arguments for accurate crawling, which can be cumbersome and error-prone in production setups.

Inaccuracies Without Agent

Without the -javaagent, strategies like Unsafe may fail for records or hidden classes (e.g., lambdas), and Specification has known issues with certain JVM arguments, reducing reliability.

Frequently Asked Questions

jamm

What is jamm?

Overview

Use Cases

Best For

Not Ideal For

Pros & Cons

Pros

Cons

Frequently Asked Questions

Related Projects

Found a gem we're missing?

jamm

What is jamm?

Overview

Use Cases

Best For

Not Ideal For

Pros & Cons

Pros

Cons

Frequently Asked Questions

Related Projects

Found a gem we're missing?