Skip to content
datastudyguides

AI-200 · 2.3.1 · Develop AI solutions by using Azure data management services

Redis cache-aside

Implement Azure Managed Redis data operations, including caching, expiration, and invalidation

Check Redis first, load and cache on a miss, set a , and invalidate only after the source database has been updated.

Read through the cache

Redis cache-aside pattern showing cache hits, cache misses that load from the database and set a TTL, and the write path that updates the database before invalidating the cached key.

In , the application checks Redis first. On a miss, read the source database, return the value, and cache a copy with a .

value = r.get(key)
if value is None:
    value = load_from_database()
    r.set(key, value, ex=300)
return value

Invalidate after the source changes

Write the source of truth first, then delete the cached key.

update_database(value)
r.delete(key)

Deleting first creates a race: another request can reload the old database value and put stale data back into Redis before the database update finishes.

Expiration is part of the design

Give cached keys a TTL so forgotten entries disappear. With an that only considers expiring keys, keys without a TTL are protected from that eviction pool.

Keep Redis operations cheap

Keep values reasonably small, use pipelining when several operations can be sent together, and avoid broad key scans such as KEYS. Azure Managed Redis uses TLS and its current endpoint port is 10000.

Key points

  • : read Redis first, then fill the cache after a miss
  • Expiration: set a when caching values
  • Write path: update the source first, then delete the cached key
  • : deleting first can reintroduce stale data during a race
  • Efficiency: small values, pipelining, and no broad `KEYS` scans
  • Connection: Azure Managed Redis uses TLS on port 10000

Click an underlined term for a one-line explanation, or see them all in the glossary .

Exam trap

Assuming the cache cleans itself: with the default volatile-lru policy, keys without a are never evicted and memory only grows.

Sources · 4 Microsoft Learn pages Show ▾

Checked on October 4, 2026. Report a problem

Go deeper on Microsoft Learn

Checked against Microsoft Learn on October 4, 2026. Report a problem with this card

How well do you know this?