Optimizing Performance

1. Using Streams Over Buffers

Approach	Memory
Read entire file into Buffer	O(file size)
Pipe streams	O(highWaterMark)
Use case	Large files, network proxying, transforms

Warning: fs.readFileSync, crypto.pbkdf2Sync, large JSON.parse all block the event loop. Never call them in request handlers.

Workload	Recommended
Image / PDF processing	Worker pool (piscina)
Crypto / hashing	Async crypto APIs first; pool for batch
JSON parsing > few MB	Stream parser or worker

Tip	Effect
Reuse buffers	Avoid alloc churn
Avoid hidden class transitions	Initialize all properties in constructor
Tune `--max-old-space-size`	Match container memory
Avoid huge global arrays	Forces major GC

Technique	Detail
Stream don't buffer	Process row-by-row
Use typed arrays	Compact numeric data
Drop unused require'd modules	Especially heavy CLIs
Use `node:sea`	Single-executable, smaller startup

Protocol	Wins
HTTP/2	Multiplexing, header compression
HTTP/3 (QUIC)	UDP-based, no head-of-line blocking v22+ experimental

import compression from "compression";
app.use(compression({ threshold: 1024 }));

Tool	Use
`fast-json-stringify`	Schema-based serialize
`simdjson` bindings	SIMD parsing
Stream parsers	`stream-json` for huge payloads

Example	Why
`sharp`	libvips bindings, faster than pure JS
`argon2` / `bcrypt`	Native CPU-bound work
Caveat	Build/install cost, ABI compatibility