Cluster Modes

Endpoints

Type	Endpoint
Single Node Redis (cluster mode disabled)	One endpoint for both reads and writes
Multi-Node Redis (cluster mode disabled)	Primary Endpoint always connects to primary node, use this for all writes Node endpoint points to a specific node, use only for reads
Multi-Node Redis (cluster mode enabled)	Configuration Endpoint provides information on Primary and Read/Node endpoints for each shard

Caching Strategies

Lazy Loading

Advantages

• Only requested data is loaded in Cache, no superfluous data
• Node failure is not fatal, although with increased latency and load on back end data store

Disadvantages

• Cache miss penalty
• Stale Data, if data only loaded in cache with the initial look up it will turn stale on next update to its values in back end data store

Write Through

Adds or updates data in cache with every update to back end data store.

Advantages

• Data is never stale
• Read latency is minimized at the cost of an increase in Write Latency

Disadvantages

• Missing Data if not combined with Lazy Loading
• Superfluous data, can be addressed with TTL

TTL

An integer value that specifies the number of seconds/milliseconds until key expires

Background Write Process and Memory Usage

Set reserved-memory-percent to 25 for Redis version 2.8.22 and later