Synchronization Benchmarks

Measure synchronization throughput, latency, batching, and database impact across representative workloads.

​

Sync & Turso performance benchmarks

Track these on a release branch before tagging. Use a Release build of the host app or a dedicated benchmark target (never --release for the Rust engine rebuild unless CI requires it).

​

Methodology

• Device: M-series Mac or recent iPhone simulator/device
• Dataset: N notes with title/body/updatedAt + optional embedding column
• Metrics: median of 5 runs after one warm-up

​

Benchmarks

Measured numbers are not yet populated. Fill these tables during the v1.0 RC and attach Instruments traces for regressions.

​

1. drainCDC

N local writes	Limit	Median ms	Notes
100	500
500	500
2_000	500 (loop)		multiple drain passes

let t0 = ContinuousClock.now
_ = try engine.drainCDC(limit: 500)
let ms = (ContinuousClock.now - t0).components.attoseconds // convert

​

2. Sync round-trip (simulated, no network)

Scenario	Median ms
A insert → makeRecord → B applyRemoteRecord
A delete → B applyRemoteDeletion
100 row batch apply

​

3. writeConcurrent (MVCC)

Concurrent writers	Rows each	Median ms	Conflicts/retries
1	100
2	100
4	100

let db = try BoutiqueDB(url: url, concurrentWrites: true)
// time writeConcurrent inserts

​

4. LiveQuery refresh after write

Rows in table	Median refresh ms
100	target < 500
1_000

​

Targets (v1.0 aspirational)

Op	Target
drainCDC 500	< 50 ms on M-series
Simulated record apply	< 5 ms
LiveQuery after local write	< 500 ms
writeConcurrent single writer 100 rows	competitive with plain write

Fill tables during RC; attach Instruments traces for regressions.