I’ve been working on AnuDB, a lightweight embedded key-value database backed by RocksDB, optimized for low-power devices like the Raspberry Pi.

It's designed as an alternative to SQLite in scenarios where high concurrency and write-heavy workloads matter.

I benchmarked AnuDB vs SQLite (with WAL mode) on a Raspberry Pi (ARMv7). Both were cross-compiled. Here are the results:

Benchmark: Operations per second (Ops/s)

Insert: AnuDB 448 | SQLite 839

Query: AnuDB 55 | SQLite 30

Update: AnuDB 408 | SQLite 600

Delete: AnuDB 556 | SQLite 1942

Parallel (10 threads): AnuDB 413 | SQLite 1.5

While SQLite is highly optimized for single-threaded operations, it struggles under multi-threaded writes. AnuDB, using RocksDB, handles parallel operations much better.

GitHub:

AnuDB: https://github.com/hash-anu/AnuDB

Benchmark: https://github.com/hash-anu/AnuDBBenchmark

Would love feedback on:

Use case suggestions

Benchmarking approaches

Whether this could be useful in your projects

Thanks!