Choose pull vs push model. For large scale, precompute timelines for active users (fan-out on write) and store in a sorted set. Cache pages, use ranking service, and backfill with pull for long-tail users.
Choose pull vs push model. For large scale, precompute timelines for active users (fan-out on write) and store in a sorted set. Cache pages, use ranking service, and backfill with pull for long-tail users.