Show Your Support: This site is primarily supported by advertisements. Ads are what have allowed this site to be maintained on a daily basis for the past 18+ years. We do our best to ensure only clean, relevant ads are shown, when any nasty ads are detected, we work to remove them ASAP. If you would like to view the site without ads while still supporting our work, please consider our ad-free Phoronix Premium.
Faster Ceph With Linux 5.16 Now That Async Dirops Have Been Flipped On
Last year with Linux 5.7 the Ceph file-system driver added the feature of async directory operations "dirops" and now finally with Linux 5.16 that is being enabled by default. After lots of testing this summer on Ceph's async dirops code, the developers are comfortable enabling it by default rather than requiring the "nowsync" mount option.
There is this USENIX presentation by Red Hat's Jeff Layton and Patrick Donnelly that covers the asynchronous directory operations for Ceph. Their numbers show removing a directory going from 10.3 seconds to just 0.3 seconds with async dirops, creating many files going from 11.3 seconds to 5.5 seconds, and other performance advantages to async directory operations.
So now with Linux 5.16 this behavior is the default while "wsync" has been added as a mount option for those preferring the old behavior or debugging purposes.
The Ceph pull request for Linux 5.16 sums up the async dirops as "greatly speeds up things like rm, tar, and rsync."