sled likes eating data! it's pre-alpha

A pre-alpha modern embedded database.

extern crate sled;

use sled::{ConfigBuilder, Tree};

let config = ConfigBuilder::new()
  .path(path)
  .build();

let tree = Tree::start(config).unwrap();

// set and get
tree.set(k, v1);
assert_eq!(tree.get(&k), Ok(Some(v1)));

// compare and swap
tree.cas(k, Some(v1), Some(v2));

// scan forward
let mut iter = tree.scan(k);
assert_eq!(iter.next(), Some(Ok((k, v2))));
assert_eq!(iter.next(), None);

// deletion
tree.del(&k);

features

• ordered map API
• fully atomic single-key operations, supports CAS
• zstd compression (use the zstd build feature)
• cpu-scalable lock-free implementation
• SSD-optimized log-structured storage

goals

1. don't make the user think. the interface should be obvious.
2. don't surprise users with performance traps.
3. don't wake up operators. bring reliability techniques from academia into real-world practice.
4. don't use so much electricity. our data structures should play to modern hardware's strengths.

plans

• beat LSM trees for reads and traditional B+ trees for writes
• MVCC, transactions, merge operators and snapshots provided via a higher-level Db versioned-key interface
• custom merge operators a la RocksDB
• form the iron core of a linearizable store and a flexible location-agnostic store
• SQLite, MySQL, Postgres back-end plugin support
• forward-compatible binary format
• dynamic learned/probabilistic index support
• bindings for other languages

warnings

• quite young, should be considered unstable for the time being
• the C API is likely to change rapidly
• the on-disk format is going to change in non-forward compatible ways before the 1.0.0 release! after that, we will always support forward migrations.
• has not yet received much attention for performance tuning, it has an extremely high theoretical performance but there is a bit of tuning to get there. currently only around 200k operations per second with mixed workloads, and 7 million/s for read-only workloads on tiny keys. this will be improving dramatically soon!

contribution welcome!

want to help advance the state of the art in open source embedded databases? check out CONTRIBUTING.md!

architecture

lock-free tree on a lock-free pagecache on a lock-free log. the pagecache scatters partial page fragments across the log, rather than rewriting entire pages at a time as B+ trees for spinning disks historically have. on page reads, we concurrently scatter-gather reads across the log to materialize the page from its fragments.

References

• The Bw-Tree: A B-tree for New Hardware Platforms
• LLAMA: A Cache/Storage Subsystem for Modern Hardware
• Deuteronomy: Transaction Support for Cloud Data
• The Design and Implementation of a Log-Structured File System