Updating VM Memory Usage Configs

Luke Hebert

January 24, 2022 - 2 minutes read - 350 words

Linux’s virtual memory subsystem makes a lot of decisions behind the scenes about when to reclaim page cache, when to flush dirty pages, and when to swap anonymous memory. If your workload is I/O-heavy or memory-sensitive, the defaults may not be ideal. Three kernel parameters give you direct control over that behavior.

Parameters

Parameter	Default	What it controls
`vm.vfs_cache_pressure`	100	How aggressively the kernel reclaims memory used for directory and inode caches. Lower values keep those caches in memory longer.
`vm.dirty_ratio`	20	The percentage of total RAM that may be dirty (waiting to be written to disk) before the kernel forces a writeback.
`vm.swappiness`	60	How strongly the kernel prefers to swap anonymous pages versus reclaiming page cache. Lower values favor keeping anonymous memory in RAM.

Applying the Settings

Add the following lines to /etc/sysctl.conf or, preferably, a dedicated file under /etc/sysctl.d/ (for example /etc/sysctl.d/99-vm-tuning.conf):

vm.vfs_cache_pressure = 100
vm.dirty_ratio = 20
vm.swappiness = 50

Then apply them immediately without rebooting:

sudo sysctl -p /etc/sysctl.d/99-vm-tuning.conf

Verifying

Confirm the values took effect:

sysctl vm.vfs_cache_pressure vm.dirty_ratio vm.swappiness

You should see output like:

vm.vfs_cache_pressure = 100
vm.dirty_ratio = 20
vm.swappiness = 50

When to Adjust

Lower vm.swappiness (e.g., 10 or 1): Useful on database servers or any workload where swapping causes noticeable latency. The kernel will reclaim page cache before touching anonymous memory.
Raise vm.dirty_ratio (e.g., 40): Helps bursty write workloads by allowing more data to sit in memory before a forced flush, reducing I/O stalls.
Lower vm.vfs_cache_pressure (e.g., 50): Beneficial when your application repeatedly accesses the same set of files and directories, since the kernel will hold onto inode and dentry caches longer.

Note

These settings are system-wide. Test them in a staging environment first, especially if you are running a production database or a latency-sensitive service. A value that helps one workload can hurt another.

Reverting

If the tuning causes unexpected behavior, restore the defaults:

sudo sysctl -w vm.vfs_cache_pressure=100
sudo sysctl -w vm.dirty_ratio=20
sudo sysctl -w vm.swappiness=60

Then remove or comment out the lines in your sysctl file so they do not reapply on the next boot.

SysAdmin