What is the fundamental architectural upgrade of Linux Cgroups v2 over Cgroups v1?

Cgroups v1 had multiple orthogonal hierarchies where different resource controllers (cpu, memory, blkio, net_cls) operated independently without shared accounting. This made coordinated resource management impossible (for example, buffered writeback IO could not be attributed to the originating memory cgroup). Cgroups v2 introduced a unified single hierarchy where all processes belong to a single tree node, enabling coordinated writeback, unified page cache accounting, and deterministic resource isolation.

How do memory.min, memory.low, memory.high, and memory.max differ in Cgroups v2?

memory.min is hard protection: the kernel will NEVER reclaim memory below this threshold, even during system-wide OOM panic. memory.low is soft, best-effort protection: memory is preserved unless un-protected cgroups are exhausted. memory.high is the proactive throttling boundary: when exceeded, the kernel slows down the cgroups executing processes in proportion to the overage, forcing them into asynchronous direct reclamation without OOM-killing. memory.max is the hard ceiling: exceeding it invokes the OOM killer.

What is the difference between "some" and "full" metrics in Pressure Stall Information (PSI)?

Pressure Stall Information (/proc/pressure/memory) tracks the percentage of wall-clock time CPU tasks spend stalled waiting for memory (page faults, direct reclamation, or swap IO). "some" indicates the fraction of time that at least one task in the cgroup was stalled. "full" indicates the fraction of time that ALL runnable tasks were stalled simultaneously, meaning the CPU was completely starved of productive work due to memory starvation.

How do production orchestrators use PSI epoll triggers for proactive auto-scaling?

Instead of polling /proc at slow multi-second intervals, applications open /proc/pressure/memory (or cgroup.pressure), write a trigger specification (e.g. "some 150000 1000000" representing 150ms of stalls in a 1-second window), and register the descriptor with epoll(). When memory pressure spikes past the threshold, the kernel immediately wakes epoll_wait(), allowing edge proxies or orchestrators to shed load, purge caches, or horizontally scale within milliseconds before an OOM occurs.

Linux Cgroups v2 Memory & Pressure Stall Information (PSI) Studio

Verified Specification | Standardized Formulas | Instant Precision

Secure & Private (Zero Data Retention) Free Access • No Sign-Up

Linux Kernel 6.x Resource Control & PSI

Cgroups v2 Memory & PSI Architecture Studio

Simulate Linux Cgroups v2 memory thresholds and Pressure Stall Information (PSI). Model memory.min, memory.low, proactive throttling (memory.high), and epoll stall event triggers.

1. Cgroup v2 Memory Thresholds

Workload Memory Pressure (MB):

256 MB 850 MB 2,048 MB

memory.low (Soft Prot):

memory.high (Throttling):

memory.min (Hard Prot):

memory.max (OOM Limit):

2. Pressure Stall Information (PSI) Gauges

Memory PSI 'some' Stall (10s): 4.2%

Memory PSI 'full' Stall (10s): 0.0%

Kernel Status: Operating normally inside protected memory boundaries.

Cgroups v2 Memory & PSI Architecture Studio

1. Cgroup v2 Memory Thresholds

2. Pressure Stall Information (PSI) Gauges

Cgroup Memory Range Allocation

Kernel Memory Events & PSI Watcher

Production Configuration (Systemd Slice & C PSI Watcher)

Cgroups v2 Memory & PSI Architecture Studio

1. Cgroup v2 Memory Thresholds

2. Pressure Stall Information (PSI) Gauges

Cgroup Memory Range Allocation

Kernel Memory Events & PSI Watcher

Production Configuration (Systemd Slice & C PSI Watcher)

🔍 Companion Workbenches & Verification Tools

Linux Kernel HugeTLB vs THP & Memory Compaction Studio | Zero TLB Thrashing

Linux io_uring Registered Buffers & Fixed Files Architecture Studio

Linux io_uring Provided Buffer Rings Studio | io_uring_register_buf_ring Architecture

Linux CXL Memory Tiering & Page Demotion Studio | Kernel Tiered NUMA Architecture

Why Smart Users Are Switching to This Free Toolkit

The Browser Feature Most People Never Use

Stop Paying for Software You Can Get Free

File Conversion Tricks That Save Hours Every Week

This Free Tool Replaced 5 Paid Apps for Our Team

What Your Browser Can Do Without Any Extensions