역압과 메시지 일괄 처리

When a producer sends faster than a consumer or the network can handle, messages pile up in buffers, memory grows, and latency spikes. Managing this is called backpressure.

Each WebSocket has an outgoing buffer. If you keep calling send faster than the socket can flush, that buffer grows unbounded.

The browser exposes bufferedAmount: bytes queued but not yet sent. A rising value signals the consumer side is falling behind.

When the buffer is high, stop producing until it drains. Resume when it falls below a low watermark.

On Node.js, ws.send can accept a callback, and the underlying stream signals when to slow down. Respect it instead of blindly flooding.

Sending thousands of tiny messages is inefficient: each has framing overhead. Batching groups many updates into one frame, cutting overhead dramatically.

Collect messages for a short window, then flush them together on a timer.

Alternatively flush when the batch reaches a target size, capping per-message latency.

For state that changes rapidly (like a cursor position), keep only the latest value instead of every intermediate one. This is coalescing, a powerful form of batching.

Batching trades a little latency for much higher throughput. Tune the window so users do not perceive lag while you still gain efficiency.

Handle flow control well:

• Watch bufferedAmount to detect pressure
• Pause producing above a high watermark
• Batch tiny messages by time or size
• Coalesce rapidly-changing state

You learned flow control over WebSockets:

• Backpressure prevents buffer overload
• Monitor bufferedAmount and pause above a watermark
• Batch tiny messages by time or size
• Coalesce fast-changing state

Your realtime channel now stays stable under load.