Skip to content

ElastiCache SOA

Links: 114 AWS SOA Index

Redis Scaling

Cluster Mode Disabled

  • Horizontal:
    • Scale out/in by adding/removing read replicas (max. 5 replicas)
  • Vertical:
    • Scale up/down to larger/smaller node type
    • ElastiCache will internally create a new node group, then data replication and DNS update.
      • attachments/Pasted image 20230306081456.jpg

Cluster Mode Enabled

  • Two Modes:
    • Online Scaling: continue serving requests during the scaling process (no downtime, some degradation in performance)
    • Offline Scaling: unable to serve requests during the scaling process (backup and restore).
      • Additional configurations supported (change node type, upgrade engine version, etc)
  • Horizontal: (Resharding and Shard Rebalancing)
    • Resharding: scale out/in by adding/removing shards
    • Shard Rebalancing: equally distribute the keyspaces among the shards as possible
    • Supports Online and Offline Scaling
  • Vertical: (change read/write capacity)
    • Scale up/down to larger/smaller node type
    • Supports Online Scaling

Redis Metrics to Monitor

  • Evictions: the number of non-expired items the cache evicted to allow space for new writes (memory is overfilled). Solution:
    • Choose an eviction policy to evict expired items (e.g., evict least recently used (LRU) items).
    • Scale up to larger node type (more memory) or scale out by adding more nodes.
  • CPUUtilisation: monitor CPU utilisation for the entire host
    • Solution: scale up to larger node type (more memory) or scale out by adding more nodes.
  • SwapUsage: should not exceed 50 MB
    • Solution: verify that you have configured enough reserved memory.
  • CurrConnections: the number of concurrent and active connections
    • Solution: investigate application behaviour to address the issue
  • DatabaseMemoryUsagePercentage: the percentage of memory utilisation
  • NetworkBytesIn/Out & NetworkPacketsIn/Out
  • ReplicationBytes: the volume of data being replicated
  • ReplicationLag: how far behind the replica is from the primary node

Memcached Scaling

  • Horizontal:
    • Add/remove nodes from the cluster
    • Auto-discovery allow your app to find nodes
  • Vertical:
    • Scale up/down to larger/smaller node type
    • Scale up process:
      • Create a new cluster with the new node type
      • Update your application to use the new cluster's endpoints
      • Delete the old cluster
      • attachments/Pasted image 20230306082429.jpg
    • Memcached clusters/nodes start out empty since there is no replication.
An application that uses an Amazon ElastiCache Memcached cluster is receiving a larger increase in traffic. A SysOps Administrator needs to use a larger instance type with more memory. What does the Administrator need to do to implement this change?

Create a new cache cluster with a new node type using the CreateCacheCluster API.

With Amazon ElastiCache Memcached engine you CANNOT modify the node type. The way to scale up is to create a new cluster and specify the new node type. You can then update the endpoint configuration in your application to point to the new endpoints and then delete the old cache cluster.

Memcached Auto Discovery

  • Auto Discovery automatically identifies all of the nodes
  • All the cache nodes in the cluster maintain a list of metadata about all other nodes.
  • This is seamless from a client perspective

Memcached Metrics

  • Same as Redis
  • FreeableMemory: amount of free memory on the host.

Last updated: 2023-03-14