Performance Benchmarks

FLASH is tuned for honest encrypted throughput — batch writes, balanced durability, and deferred Merkle rebuilds on bulk paths.

---

Benchmark Results (v1.2.5)

Measured with npm run benchmark on Apple Silicon (Node 20+):

===============================================================
⚡ FLASH High-Performance Engine Benchmark ⚡
===============================================================

🔧 Engine profile: durability=balanced, memtable=4MB, worker flush=on

📊 1. Binary Document Serialization (10,000 Operations)
   - Traditional JSON parse + field lookup:   ~230,000 ops/sec
   - FlashBinary Zero-Copy O(1) field lookup: ~400,000 ops/sec
   🚀 FlashBinary Speedup: ~1.7x vs JSON

🔐 2. Cryptographic Throughput (5,000 Operations)
   - AES-256-GCM + HMAC Blind Indexing: ~26,000 ops/sec

💾 3. End-to-End Database Engine Throughput (2,000 Documents)
   - insertOne (balanced WAL batching):     ~330 ops/sec
   - insertMany turbo batch (2,000 docs):   ~2,650 ops/sec
   - Encrypted Blind Index Point Reads:     ~6,000 ops/sec

🔒 Tamper-Proof Merkle State Root: (computed via refreshMerkleRoot)
===============================================================

INFO

1M+ ops/sec refers to FlashBinary zero-copy field lookups — not encrypted end-to-end writes. Always check the benchmark section that matches your workload.

---

Workload Guide

Workload	Recommended API	Durability
Single encrypted inserts	insertOne	balanced (default)
Bulk import / seed data	insertMany	balanced or throughput
Audit / financial log	insertOne	strict
Benchmark / disposable data	insertMany	throughput + close()

See Engine Options for configuration.

---

Running Locally

npm run benchmark

Results vary by CPU, disk (SSD vs HDD), and Node version.

---

What Changed in v1.2.5

Setting	Before	After
Memtable threshold	64 KB	4 MB
Default durability	fsync every write	balanced (batch 64 ops / 25 ms)
Bulk oplog	per-doc append	appendBatch
L0 compaction trigger	4 SSTables	8 SSTables