Libuv has introduced support for io_uring file operations, which provides a significant increase in throughput on Linux systems. The io_uring approach is used if the kernel is new enough, otherwise, the traditional thread pool method would be used. All file operations are run on the thread pool by default, and the thread pool size can be adjusted as needed. The new io_uring integration also comes with new metrics APIs, as well as changes to the thread pool threads stack size and timer execution order. This work was sponsored by ISC, the Internet Systems Consortium.
https://github.com/libuv/libuv/pull/3952