Caching & CDN
Azure Managed Redis · Front Door caching · APIM cache
Practice questions for this topic →
Options
| Service | Use for | Status |
|---|---|---|
| Azure Managed Redis | App data and session cache; built on Redis Enterprise; clustered by default; zone redundant by default with HA; Microsoft Entra ID authentication; data persistence GA | Current |
| Managed Redis performance tiers | Memory Optimized (8:1 memory-to-vCPU), Balanced (4:1), Compute Optimized (2:1), all in memory; Flash Optimized stores data across memory and NVMe and lacks RediSearch, RedisBloom, RedisTimeSeries and active geo-replication (Microsoft Learn and the Azure pricing page) | Current |
| Azure Cache for Redis | Same use cases | Basic/Standard/Premium retire Sep 30, 2028 (disabled Oct 1, 2028). Enterprise and Enterprise Flash retire March 2027 |
| Front Door Standard/Premium | Global edge caching, routing and WAF (managed WAF rule sets and Private Link origins need Premium) | Current |
| Front Door (classic) / CDN Standard from Microsoft (classic) | Edge caching | Retire Mar 31, 2027 / Sep 30, 2027 |
| APIM caching policy | Cache API responses (built-in or external Redis) | Current |
Redis retirement and migration details
| Item | Detail |
|---|---|
| Enterprise / Enterprise Flash | Retire March 31, 2027 (Microsoft's update title says March 30). New creation blocked from Apr 1, 2026. Remaining instances are migrated to Azure Managed Redis automatically after Apr 1, 2027 |
| Basic / Standard / Premium | Retire Sep 30, 2028. From Apr 1, 2026 new customers can't create caches; existing customers can keep creating until retirement. Azure Government and China clouds have no creation block |
| Connection changes | Hostname suffix becomes <region>.redis.azure.net; port 10000 for TLS or plaintext (one mode per cache, set at creation), instead of 6380 / 6379 |
| Clustering | Every instance is clustered internally. Policies: OSS (Redis Cluster API; clients reach shards directly; higher maximum throughput), Enterprise (one endpoint, a proxy routes to shards), Non-clustered (no sharding). Chosen when you create the cache. Moving from a non-clustered topology: consider OSS; use non-clustered only if the app supports neither OSS nor Enterprise |
| Cost options | Non-HA mode (dev/test) halves the instance cost; Enterprise had no non-HA option |
| Not supported | Scaling down/in and VNet injection (per Microsoft's migration reference); Flash Optimized has no RediSearch, RedisBloom, RedisTimeSeries or active geo-replication |
Microsoft's pages give both March 30 and March 31, 2027 for the Enterprise retirement. If a question asks the exact day, expect the timeline (March 2027) rather than a trick on one day.
Why it's faster
- Community Redis (Basic/Standard/Premium) is single-threaded, one process per node. Azure Managed Redis runs several shards per node in parallel, which improves vCPU use and enables active geo-replication.
Exam Tips
Tick each tip once you can answer it without looking.
“Reduce database load with low-latency reads / store session state” → Azure Managed Redis (cache-aside pattern). Trap: Azure Cache for Redis is retiring.
“Cache must be writable in multiple regions” → Azure Managed Redis active geo-replication. Trap: Not offered on the Flash Optimized tier.
“Need RediSearch, RedisBloom or RedisTimeSeries modules” → Azure Managed Redis Memory Optimized, Balanced or Compute Optimized. Trap: Flash Optimized lacks RediSearch, RedisBloom, RedisTimeSeries and active geo-replication (RedisJSON is available). Basic, Standard and Premium have no modules.
“Cache authentication must use Entra ID instead of access keys” → Azure Managed Redis.
“Serve static content from edge locations worldwide” → Front Door Standard/Premium. Trap: CDN classic and Front Door classic are retiring.
“Edge caching, but the origin must not be publicly reachable” → Front Door Premium with Private Link origin.
“Cache backend API responses without changing the app” → APIM cache-lookup / cache-store policies.
“Azure Cache for Redis Enterprise or Enterprise Flash instance isn't migrated before the retirement” → Remaining instances are migrated to Azure Managed Redis automatically after April 1, 2027 (retirement March 2027). Trap: Basic, Standard and Premium retire later (Sep 30, 2028) and are disabled from Oct 1, 2028.
“App must reconnect after moving to Azure Managed Redis” → New hostname (
“Cheapest Redis for dev/test” → Azure Managed Redis non-HA mode (halves the cost). Trap: No high availability; the Enterprise tier had no non-HA option.
“Why Azure Managed Redis outperforms the older Basic/Standard/Premium tiers” → Redis Enterprise stack runs several shards per node (multi-threaded). Trap: Community Redis is single-threaded, one process per node.
“New Basic/Standard/Premium cache after April 1, 2026” → Blocked for new customers; existing customers can keep creating until retirement. Trap: No creation block before retirement in Azure Government or China clouds.
“Azure Managed Redis tier ratios” → Memory Optimized 8:1, Balanced 4:1, Compute Optimized 2:1 memory-to-vCPU; Flash Optimized uses memory plus NVMe. Trap: Compute Optimized for maximum throughput; Memory Optimized for large data with moderate throughput.
“Migrating a non-clustered Redis cache to Azure Managed Redis” → Consider the OSS clustering policy for better performance; use non-clustered only if the app can't support OSS or Enterprise. Trap: Every Managed Redis instance is clustered internally; the policy decides what clients see.
“Redis clustering policy choices” → OSS (Redis Cluster API, clients reach shards directly, higher throughput), Enterprise (one endpoint via a proxy), Non-clustered. Trap: Chosen when you create the cache.
“Azure Cache for Redis Enterprise to Azure Managed Redis in date order” → Creation blocked Apr 1, 2026; tiers retire Mar 2027; automatic migration after Apr 1, 2027. Trap: Basic, Standard and Premium retire Sep 30, 2028 and are disabled Oct 1, 2028.