io-uring

PostgreSQL 19 introduces kernel asynchronous reads via io_uring, enabling direct asynchronous buffered I/O for improved performance without dedicated worker processes.

Silk is a new stackful-fiber library and scheduler for ClickHouse, designed to improve asynchronous I/O performance with NUMA-aware work stealing, io_uring, and zero heap allocation in the hot path. It targets I/O-bound components like distributed cache, object storage, and network I/O.

The author shares a technical deep dive into saturating NIC and disk bandwidth using RDMA and io_uring, highlighting how an AI agent achieved the optimization with a flawed explanation, then details the actual debugging process.

A technical comparison of epoll and io_uring in Linux, explaining their architectures and performance characteristics for asynchronous I/O.

The article explores using mmap to create a magic ring buffer where two contiguous virtual memory ranges map the same physical memory, and tests its compatibility with io_uring's registered buffers, finding that it works as expected.

zeroserve now supports Caddy-compatible mode, achieving 3x throughput and 70% lower latency by JIT-compiling Caddyfiles to eBPF and native machine code.

The article details the implementation of automatic buffer management using io_uring's buffer rings in UringMachine, a Ruby gem for asynchronous I/O. It explains how buffer rings enable efficient multishot read/recv operations by allowing the kernel to use application-provided buffers.

A technical article comparing three approaches to serving files over HTTP: synchronous thread-per-request, epoll-based asynchronous I/O, and io_uring, with code examples in C.

Silk is an open-source cooperative fiber scheduler for Linux with per-CPU scheduler threads, io_uring integration, and topology-aware work-stealing, designed for high concurrency with low overhead.

Zig 0.16 ships a new std.Io interface for cross-platform I/O. The library zio provides a full async implementation using stackful coroutines and OS-level async APIs, enabling efficient concurrent tasks without thread-per-task overhead.

A local privilege escalation exploit in the Linux kernel's io_uring subsystem via a zero-copy receive freelist bug.