Skip to main content

Storage engine research

The storage engine behind Api-DB has not been chosen yet. We are researching candidates now and will commit once they have been benchmarked against Api-DB's workloads.

Candidates​

Rust-native engine​

A memory-safe, embeddable database engine written in Rust, offering predictable latency and low operating cost.

Object-storage (S3) backed​

Records persisted to S3-compatible object storage, with compute separated from data for durability and near-unlimited scale.

Hybrid architectures​

A fast transactional tier in front of an object-storage foundation, combining low-latency writes with inexpensive durable storage.

How we are choosing​

Each candidate is evaluated on:

  • Durability — no acknowledged write is ever lost.
  • Throughput and latency — for both record-at-a-time API traffic and bulk loads.
  • Schema evolution — how cheaply a published Apiome schema change can be applied to stored data.
  • Multi-tenancy — isolation and fair resource sharing across tenants.
  • Operating cost — at prototype scale and at production volumes.

The outcome of this research will be published here.