Aether: High-performance, formally-verified storage engine written in Rust
2 days ago
- #Performance
- #Rust
- #Database Storage Engine
- Rust is the primary language used at 93.3%, with C, Clojure, Shell, Python, and others making up the remainder.
- A refactor was performed on Level<I> to store runs behind Arc<Run<I>>, enabling cheap cloning for the optional left-right read path.
- Key changes include updates in level.rs, the creation of lr_levels.rs, modifications in mod.rs, query.rs, and adaptive.rs, with added unit tests and benchmarks.
- Aether DB is a high-performance, formally-verified database storage engine written in Rust, featuring ACID compliance, B+ Tree, WAL, ARIES recovery, LeanStore buffer manager, multiple index types, LSM tree framework, and ACID transactions.
- It includes a unified CLI and TUI, management tools, various index types, LSM tree framework, storage engine, transaction support, formal verification, and comprehensive performance details.
- Building from source requires Rust 1.75+, with extensive testing and benchmarks available. Examples demonstrate core features, production applications, and C-compatible FFI.
- Configuration allows tuning buffer pool and index behavior. Documentation covers getting started, performance, architecture, development, and includes a roadmap from v1.0.0 to v2.0.0.
- The project is inspired by several influential systems, licensed under MIT, with contributions welcome and acknowledgments given.