Amazon ElastiCache

• ElastiCache is a service offering managed Redis or Memcached instances
• Caches are in-memory databases with really high performance and low latency
• Caches help reduce load from databases for read intensive workloads
• AWS takes cares of OS maintenance/patching, optimizations, setup, configuration, monitoring, failure recovery and backups for cache instances

ElastiCache Solution Architectures

• Database Cache:
  • Applications query ElastiCache. If the data is not available, it will be retrieved from RDS and stored in ElastiCache as well
  • Used to relieve load from RDS
  • The cache must have an invalidation strategy to make sure only the most current data is stored
• User Session Store for Stateful Applications:
  • User logs into an application
  • The application writes the session data into ElastiCache
  • In case the user hits another instance running the application, the application can retrieve the session from the cache

Comparison between Redis and Memcached

Redis	Memcached
Multi AZ with automatic failover	Uses Multi-node for partitioning of data (sharding)
Read Replicas to scale reads and have high availability	No high availability
Data durability using AOF persistance	No persistance of data
Backup and restore features	No backup and restore
Supports Sets a Sorted Sets	Has a multi-threaded architecture

ElastiCache Replication

• Cluster Mode Disabled:
  • One primary node, up to 5 replicas
  • Replication is asynchronous
  • Primary node is used for reads and writes
  • Replicas can be used for reads only
  • We have on shard, all the nodes have all the data
  • We have guarding against data loss if a node fails
  • Multi-AZ is enabled by default for failover
  • Helpful to scale read performance
  • We have 2 kinds of scaling:
    • Horizontal: Scale out/in by adding or removing read replicas (max 5 read replicas)
    • Vertical:
      • Scale up/down by having larger or a smaller node type
      • ElastiCache will internally create a new node groups, then data replication will happen, finally the DNS will be updated
• Cluster Mode Enabled:
  • Data is partitioned across multiple shards
  • The data is divided between shards
  • Helpful to scale writes
  • Each shard has a primary and up to 5 replica nodes (same concept as Cluster Mode Disabled)
  • Multi-AZ capability os ON by default
  • We get up to 500 nodes per cluster:
    • 500 shards with a single master
    • 250 shards with 1 master and 1 replica
    • …
    • 83 shards with one master and 5 replicas

Redis Auto Scaling

• We can have auto scaling when we have Cluster Mode Enabled
• We can automatically increase/decrease the desired shards or replicas
• Auto scaling supports both Target Tracking and Scheduled Scaling Policies
• To connect to the cluster it is recommended to use the Cluster’s Configuration Endpoint

Redis Connection Endpoints

• Standalone Node:
  • One endpoint for read and write operations
• Cluster Mode Disabled:
  • 1 Primary Endpoint for all the operations
  • Multiple Reader Endpoints: evenly split read operations across all the read replicas
  • Node Endpoint: read operations for individual nodes
• Cluster Mode Enabled:
  • Configuration Endpoint: for all read/write operations that support Cluster Mode Enabled command
  • Node Endpoints: for read operations only