MGLRU Is A Very Enticing Enhancement For Linux In 2022
Going back a number of months Google engineers have been working to address the issue of the Linux kernel's page reclaim code being too expensive for which they devised the multi-generational LRU framework "MGLRU" and it continues being worked on with mainline ambitions.
MGLRU has yielded very promising results from servers down through Chrome OS and Android devices too. MGLRU aims to make better choices than the current kernel page reclaim code and to do so more efficiently. Previous numbers punted by Google engineers were cold start times reduced by up to 16% while enjoying fewer low-memory kills, Chrome OS saw upwards of 59% fewer out-of-memory kills and 96% fewer low-memory tab discards in its browser, and server results have been very promising too.
Sent out this past week were the MGLRU v6 patches. The latest benchmarks across Redis, PostgreSQL, MongoDB, Memcached, Hadoop, Spark, Cassandra, MariaDB, and other workloads all look very promising.
The most important information for Linux end-users if not concerned with all the technical details about Multi-Gen LRU:
This "magic" happens in just over three thousand lines of new kernel code. Hopefully MGLRU will manage to be mainlined in the near future and I'll also be working on some of my own performance tests of this exciting kernel improvement around page reclamation.
MGLRU has yielded very promising results from servers down through Chrome OS and Android devices too. MGLRU aims to make better choices than the current kernel page reclaim code and to do so more efficiently. Previous numbers punted by Google engineers were cold start times reduced by up to 16% while enjoying fewer low-memory kills, Chrome OS saw upwards of 59% fewer out-of-memory kills and 96% fewer low-memory tab discards in its browser, and server results have been very promising too.
Sent out this past week were the MGLRU v6 patches. The latest benchmarks across Redis, PostgreSQL, MongoDB, Memcached, Hadoop, Spark, Cassandra, MariaDB, and other workloads all look very promising.
The most important information for Linux end-users if not concerned with all the technical details about Multi-Gen LRU:
This "magic" happens in just over three thousand lines of new kernel code. Hopefully MGLRU will manage to be mainlined in the near future and I'll also be working on some of my own performance tests of this exciting kernel improvement around page reclamation.
14 Comments