Lesson 15 / 25

When to Use Storage and Realtime

Fit, limits and alternatives.

By · 15 years in web development

Pick the lightest tool

Store large binary files in Storage and keep only their paths in tables. Use Postgres Changes for modest volumes of row-level updates such as dashboards or task boards; for high-traffic feeds, Broadcast (optionally sent from the database) scales better because every Postgres change must be checked against RLS for each subscriber. Simple refresh-on-focus or polling is often enough when updates are rare. Connection and message limits depend on your plan, so check the docs and pricing page.

A decision sketch

Rules of thumb.

# images, PDFs, video                      -> Storage (save the path in a table)
# few users watching row edits             -> Realtime Postgres Changes
# chat, cursors, typing, high message rate  -> Realtime Broadcast
# who is online in a room                   -> Realtime Presence
# data changes rarely                       -> refetch on focus or poll

Never store files as base64 in rows

Large blobs bloat the database, backups and every query; Storage and a CDN serve them far more cheaply.

Quick check: Where should a user-uploaded video be kept?

  • In Storage, with its path saved in a table
  • As base64 in a text column
  • In a Realtime channel
  • In an environment variable
Answer

In Storage, with its path saved in a table — Keep blobs out of the database.