Zstandard Compression Algorithm

Zstandard (zstd) is a fast real-time compression algorithm developed by Meta (formerly Facebook), providing high compression ratios with extremely fast decompression. Released in 2015 by Yann Collet (also creator of LZ4), Zstandard is standardized as RFC 8478 and has been adopted across Linux kernel, databases, and container runtimes. Key features: compression ratio better than zlib/gzip on most data, approaching LZMA/xz at max levels. Decompression speeds of 1-3 GB/s, 3-5x faster than zlib, leveraging SIMD instructions and branch-free code paths. Compression levels 1-22, with level 3 default and negative levels (-1 to -7 in --ultra mode) for fastest compression. Long-distance matching (--long) with 128 MB sliding window for large files. Dictionary compression for small files (under 100 KB) using trained dictionaries created by zstd --train, achieving dramatic ratio improvements on JSON records and network packets. Frame format with content size, xxhash checksum, and dictionary ID. Streaming API for arbitrary-size data. Multi-threaded compression using multiple CPU cores. zstd CLI tool with gzip-like interface. Language bindings for Python, Java, Go, Rust, and JavaScript. Adopted in Linux kernel (Btrfs 5.1+, SquashFS 4.4+), Hadoop 3.3+, Kafka, MongoDB, PostgreSQL (TOAST compression since PG14), Docker (container layer compression), and Homebrew. Latest stable version 1.5.5 released in 2023. Open source under BSD.