A real-time data streaming service for high-throughput, ordered data streams. It is typically used as an ingestion buffer, protecting the later layers from bursts of data

Key Features:

  • Pull-based: Consumers read from shards in order.
  • Preserves order per shard.
  • Scales to gigabytes per second.
  • Replayable: Consumers can re-read past data within a window (default 24 hours, up to 7 days).
  • Supports multiple parallel consumers (multi-subscriber).

Use Case:

  • Real-time analytics, stream processing, machine learning.
  • Handling logs, metrics, clickstreams, IoT events.