5 releases
0.1.4 | Apr 10, 2024 |
---|---|
0.1.3 | Mar 4, 2024 |
0.1.2 | Jan 3, 2024 |
0.1.1 | Jan 1, 2024 |
0.1.0 | Nov 30, 2023 |
#85 in Database implementations
50KB
1.5K
SLoC
GhalaDb
A key value datastore that implements keys and values separation inspired by the WiscKey paper.
GhalaDb implements a SSD-conscious data layout by decoupling the storage of keys from values. An in-memory tree stores the keys along with pointers to the values, while the values are stored in a separate log file. This significantly reduces write amplification during ingestion, while facilitating faster data loading.
Since GhalaDb keeps all its keys and data pointers in memory, it is suitable for applications that have small-sized keys.
use ghaladb::{GhalaDb, GhalaDbResult};
fn main() -> GhalaDbResult<()> {
let mut db = GhalaDb::new("/tmp/ghaladb", None)?;
let key = "king".to_owned();
let val = "queen".to_owned();
db.put(&key, &val)?;
assert_eq!(db.get(&key)?.unwrap(), val);
Ok(())
}
References
- https://arxiv.org/pdf/1812.07527.pdf LSM-based Storage Techniques: A Survey
- http://www.benstopford.com/2015/02/14/log-structured-merge-trees/ Log Structured Merge Trees
- https://dl.acm.org/doi/pdf/10.1145/3514221.3522563 - Dissecting, Designing, and Optimizing LSM-based Data Stores
- https://www.infoq.com/articles/API-Design-Joshua-Bloch/ Joshua Bloch: Bumper-Sticker API Design
Dependencies
~2MB
~46K SLoC