10 Key Enhancements in the 2026 DAMON Update: What Linux Developers Need to Know

At the 2026 Linux Storage, Filesystem, Memory Management, and BPF Summit, SeongJae Park, the creator of the DAMON (Data Access Monitoring) subsystem, delivered a comprehensive update on the latest developments. DAMON, which provides user-space monitoring and management of system memory, has seen rapid evolution. This article highlights 10 major enhancements introduced in this update—from tiering support to new data attributes monitoring—that are reshaping how Linux manages memory.

1. Tiering Support for Heterogeneous Memory

One of the most significant additions is formal tiering support. DAMON now enables monitoring and management across multiple memory tiers, such as DRAM, persistent memory (PMEM), and CXL-attached memory. Administrators can define policies to automatically migrate hot pages to faster tiers and cold pages to slower ones. This feature leverages DAMON's existing access detection to trigger page migrations via mglru or the legacy reclaim mechanism. The result is improved performance and cost efficiency for systems with diverse memory types, particularly in data centers and high-performance computing.

2. Fine-Grained Data Attributes Monitoring

DAMON has expanded beyond simple access frequency to monitor a richer set of data attributes. You can now track per-page metrics like last access time, access count, and even the type of operation (read vs. write). This is achieved through new hooks in the memory management subsystem that record metadata without significant overhead. For example, a developer can use these attributes to identify “write-hot” pages for SSD caching or “read-intensive” regions for prefetching. The monitoring API has been extended to return these attributes in user-space bitmaps, making it easier to build custom memory optimization tools.

3. Transparent Huge Pages (THP) Integration

DAMON now offers transparent huge pages-aware monitoring. Earlier versions treated each base page (4 KB) independently, but large page sizes (e.g., 2 MB) are common in modern workloads. The updated subsystem can monitor huge pages as atomic units, breaking them down into smaller regions only when needed. This integration reduces metadata overhead and provides more accurate access patterns for applications like databases and virtual machines. Additionally, DAMON can trigger THP collapsing or splitting based on observed access patterns, improving performance without manual tuning.

4. Low-Overhead Adaptive Sampling

A perennial challenge for memory monitoring is balancing accuracy with CPU overhead. The 2026 DAMON update introduces adaptive sampling intervals that dynamically adjust based on observed access rates. When memory regions are idle, the sampling frequency drops to conserve resources; during heavy access, it ramps up to capture fine-grained behavior. The algorithm uses a PID controller to maintain a target overhead percentage, configurable via sysfs. Early benchmarks show a 30–50% reduction in monitoring overhead for typical server workloads while preserving high accuracy for critical memory regions.

5. User-Space API Enhancements

The user-space interface has been overhauled for flexibility and ease of use. A new sysfs-based control set allows administrators to configure monitoring regions, attributes, and actions without recompiling the kernel. For programmatic control, the existing ioctl interface now supports batched operations, reducing context-switch overhead when managing many monitoring targets. Additionally, a new debugfs file exposes real-time access statistics in human-readable format, aiding debugging. These changes make DAMON more accessible for both system integrators and developers building custom memory management solutions.

6. Proactive Reclaim and Memory Pressure Feedback

DAMON can now drive proactive memory reclaim by providing early feedback to the kernel’s memory pressure algorithm. Instead of waiting for kswapd to detect high pressure, DAMON’s monitoring predicts cold pages and prompts their reclaim before system performance degrades. This is implemented via a new DAMON_RECLAIM action that preemptively sends pages to swap or discards them based on policy. Coupled with the tiering support, proactive reclaim ensures that under memory pressure, the system prioritizes keeping hot data in the fastest tier, improving overall responsiveness.

7. Per-Process and Per-Cgroup Monitoring

Memory optimization often requires per-application granularity. This update introduces per-process and per-control group (cgroup) monitoring capabilities. Users can attach DAMON to a specific PID or cgroup, and the subsystem will track only that process’s address space or memory usage within the cgroup. The access patterns are reported relative to the process’s virtual memory layout, simplifying interpretation. This is particularly valuable for container environments where resource isolation is critical, allowing operators to fine-tune memory for each workload without affecting others.

8. Integration with Memory Tiering Policies (MTP)

Building on new kernel memory tiering infrastructure, DAMON now works seamlessly with Memory Tiering Policies (MTP). Administrators can define policies—such as “promote after 10 accesses” or “keep in DRAM until idle for 60 seconds”—and DAMON executes them by monitoring access events and issuing migration commands. This eliminates the need for external tools to manage multi-tier memory, reducing complexity and potential race conditions. The policies are stored in a simple YAML-like format in sysfs, making them easy to version-control and audit.

9. Reduced Monitoring Overhead via Hardware Counters

For platforms that support Intel PMU or AMD IBS, DAMON can now offload access detection to hardware counters. This cuts software overhead dramatically—by up to 90% in preliminary tests—while maintaining accuracy equivalent to page-table-based monitoring. The feature is opt-in and works in combination with existing software sampling. When hardware counters are unavailable, DAMON falls back to its proven pseudo-LRU approach. This hybrid model makes DAMON viable for latency-sensitive workloads that previously avoided memory monitoring due to performance concerns.

10. Comprehensive Documentation and Test Suite

Last but not least, the 2026 update ships with overhauled documentation and a unified test suite. The kernel documentation now includes a dedicated DAMON chapter covering all features, configuration examples, and best practices. Additionally, a new set of unit tests and integration tests (under tools/testing/selftests/damon) validates each feature across different architectures and memory configurations. This ensures that DAMON remains reliable as it evolves and lowers the barrier for contributors to add new features without breaking existing functionality.

In summary, the 2026 DAMON update represents a major leap forward for Linux memory management. With support for heterogeneous memory, finer-grained monitoring, and reduced overhead, it equips system administrators and kernel developers with powerful tools to optimize performance and efficiency. These enhancements also pave the way for future innovations in tiered memory systems, proactive reclaim, and adaptive resource management. Whether you’re managing a cloud data center or a single embedded system, DAMON’s new capabilities are worth exploring.