The Stream Data Type

An append-only log in a key

A stream is a Redis key holding an append-only sequence of entries. Each entry has a unique ID and a small set of field-value pairs, like a tiny hash: orderId o-1001 status placed. IDs have the form <milliseconds>-<sequence>, such as 1727942712345-0: the first part is the Redis server time in milliseconds when the entry was added, and the sequence distinguishes entries added in the same millisecond. IDs always increase, so a stream is ordered and you can query by ID ranges, which are effectively time ranges. Unlike Pub/Sub, entries persist (in memory, and on disk if persistence is enabled) until you trim or delete them, so consumers can read history and resume where they left off. Unlike lists, reading does not remove entries, so many independent readers can consume the same stream. Internally, streams use a radix tree of compact listpacks, making them memory-efficient for many small entries.

Entries with time-based IDs

Each entry has an increasing millisecond-based ID and a few fields; readers keep their own position.

Adding and reading entries

* asks Redis to generate the ID; - and + mean the smallest and largest IDs.

XADD orders '*' orderId o-1001 customer c-42 status placed
# returns an ID such as 1727942712345-0
XADD orders '*' orderId o-1002 customer c-7 status placed

XLEN orders                         # number of entries
XRANGE orders - +                   # all entries, oldest first
XRANGE orders - + COUNT 10          # first 10
XREVRANGE orders + - COUNT 5        # newest 5
XRANGE orders 1727942700000 1727942800000   # by time range (ms)

Keep entries small

Streams shine with many small entries. Put identifiers and essential fields in the entry and keep large payloads (documents, images) in a database or object store, referenced by ID.

Quick check: What does the first part of a stream entry ID such as 1727942712345-0 represent?

• A random number
• The consumer ID
• The millisecond Unix time when the entry was added
• The entry size in bytes