SkipprDB
Very excited about this early-stage k/v, strongly consistent (read-after-write), fully serialised transactional database. There are two primary design goals:
- Full DynamoDB data API support (Scan, Query, Filter, conditionals, etc.) and full ANSI SQL support (not just existing DynamoDB PartiQL).
- Modern high performance with very tight read and write tail latencies, by optimising data structures and cluster co-ordination to make full use of the latest server hardware (inc. DMA, NUMA, CPU cache lines, 4K-aligned NVMe storage, etc.)
The design is inspired by FoundationDB, the original DynamoDB white paper, and TigerBeetle financial transaction database. Currently completing the specification design and moving onto formal spec proofs via P and TLA soon.
DesignDurabilityP specs