MGLRU Patches Picked Up By Andrew Morton's "mm-unstable" Branch Ahead Of Linux 6.1
MGLRU overhauls the Linux kernel's page reclamation code so it's smarter about page evictions and to do so in a more performant manner. MGLRU has shown to be of big advantage for Linux systems with limited amounts of memory and enhancing overall system performance. MGLRU was started by Google engineer Yu Zhao after finding the existing kernel page reclamation code to be insufficient and expensive.
MGLRU benchmarks are looking good from a variety of parties and many different workloads. After not aligning for the Linux 6.0 cycle, MGLRU is one of the features that has me already eager for Linux 6.1.
Today as part of working its way toward Linux 6.1, the MGLRU patches were picked up by Andrew Morton for his mm-unstable branch. The longtime kernel developer is the one that recently laid out the plans for getting both MGLRU and Maple Tree into Linux 6.1.
So the latest MGLRU work is in that memory management kernel branch for further testing over the coming weeks and assuming no last minute problems should be sent for mainline as part of the Linux 6.1 merge window in early October.
Those wishing to learn more about Multi-Gen LRU can see this design document as part of the queued patches.