How to Build Real-Time Live-Stream Comments: Polling, WebSocket, and SSE Compared

Background

Live‑streaming platforms need to deliver user comments to all participants with minimal latency. The following sections compare three common techniques and describe a scalable Server‑Sent Events (SSE) architecture that provides high‑availability real‑time comment delivery.

Implementation Options

1. HTTP Polling

Clients repeatedly issue GET requests (e.g., every 1 second) to fetch new comments. This pull model is simple but introduces noticeable latency because a comment is only received after the next poll. When comment traffic is low, the frequent requests waste bandwidth and server resources.

2. WebSocket

WebSocket establishes a persistent, bidirectional TCP connection. After a client sends a comment, the server pushes the comment to all other connected clients instantly and also publishes a message to a message‑queue (MQ) for asynchronous persistence.

Although this eliminates polling latency, maintaining a WebSocket connection for every viewer is costly in read‑heavy scenarios where most users only consume comments. The overhead of keeping many idle connections can outweigh the benefits.

3. Server‑Sent Events (SSE)

SSE uses a single HTTP connection to stream events from server to client (unidirectional). The server can push new comments to all listeners as soon as they arrive, while clients only need to read the stream.

Because the connection is lightweight and only one‑way, SSE is well‑suited for live rooms where comment generation is sparse but comment consumption is heavy.

Scalable SSE Architecture

Problem in a naïve cluster

When multiple SSE instances sit behind a load balancer, a client’s SSE connection may be bound to server A while the comment originates on server B. Server B cannot directly push the comment to the client connected to server A, causing missed updates.

Upgraded design

The solution separates two responsibilities:

SSE Connection Service : maintains only the long‑lived SSE streams.

Comment Service : handles comment creation, validation, and storage.

When a comment is posted, the Comment Service publishes two MQ messages:

A “push” message consumed by the SSE cluster to broadcast the comment to every SSE connection belonging to the same live‑room.

A “persist” message consumed by a background worker that writes the comment to the database.

The SSE cluster subscribes to the push MQ, receives the comment regardless of which server processed it, and forwards the comment to all connected clients (e.g., users A, B, C) in that room. This achieves real‑time delivery with high availability.

Remaining limitations

As comment volume grows, the number of MQ messages increases proportionally. High MQ traffic can become a bottleneck, potentially increasing end‑to‑end latency and reducing the perceived real‑time performance of the system.