Target tracking scaling policies - Amazon ElastiCache
Predefined metricsAuto Scaling criteria for shardsAuto Scaling considerations

Target tracking scaling policies

With target tracking scaling policies, you select a metric and set a target value. ElastiCache for Valkey and Redis OSS Auto Scaling creates and manages the CloudWatch alarms that trigger the scaling policy and calculates the scaling adjustment based on the metric and the target value. The scaling policy adds or removes shards as required to keep the metric at, or close to, the specified target value. In addition to keeping the metric close to the target value, a target tracking scaling policy also adjusts to the fluctuations in the metric due to a fluctuating load pattern and minimizes rapid fluctuations in the capacity of the fleet.

For example, consider a scaling policy that uses the predefined average ElastiCachePrimaryEngineCPUUtilization metric with configured target value. Such a policy can keep CPU utilization at, or close to the specified target value.

Predefined metrics

A predefined metric is a structure that refers to a specific name, dimension, and statistic (average) of a given CloudWatch metric. Your Auto Scaling policy defines one of the below predefined metrics for your cluster:

Predefined Metric Name CloudWatch Metric Name CloudWatch Metric Dimension Ineligible Instance Types
ElastiCachePrimaryEngineCPUUtilization

EngineCPUUtilization

ReplicationGroupId, Role = Primary

 None
ElastiCacheDatabaseCapacityUsageCountedForEvictPercentage

DatabaseCapacityUsageCountedForEvictPercentage

Valkey or Redis OSS Replication Group Metrics

 None
ElastiCacheDatabaseMemoryUsageCountedForEvictPercentage

DatabaseMemoryUsageCountedForEvictPercentage

Valkey or Redis OSS Replication Group Metrics

 R6gd

Data-tiered instance types cannot use ElastiCacheDatabaseMemoryUsageCountedForEvictPercentage, as these instance types store data in both memory and SSD. The expected use case for data-tiered instances is to have 100 percent memory usage and fill up SSD as needed.

Auto Scaling criteria for shards

When the service detects that your predefined metric is equal to or greater than the Target setting, it will increase your shards capacity automatically. ElastiCache for Valkey and Redis OSS scales out your cluster shards by a count equal to the larger of two numbers: Percent variation from Target and 20 percent of current shards. For scale-in, ElastiCache won't auto scale-in unless the overall metric value is below 75 percent of your defined Target.

For a scale out example, if you have 50 shards and

  • if your Target breaches by 30 percent, ElastiCache scales out by 30 percent, which results in 65 shards per cluster.

  • if your Target breaches by 10 percent, ElastiCache scales out by default Minimum of 20 percent, which results in 60 shards per cluster.

For a scale-in example, if you have selected a Target value of 60 percent, ElastiCache won't auto scale-in until the metric is less than or equal to 45 percent (25 percent below the Target 60 percent).

Auto Scaling considerations

Keep the following considerations in mind:

  • A target tracking scaling policy assumes that it should perform scale out when the specified metric is above the target value. You cannot use a target tracking scaling policy to scale out when the specified metric is below the target value. ElastiCache for Valkey and Redis OSS scales out shards by a minimum of 20 percent deviation of target of existing shards in the cluster.

  • A target tracking scaling policy does not perform scaling when the specified metric has insufficient data. It does not perform scale-in because it does not interpret insufficient data as low utilization.

  • You may see gaps between the target value and the actual metric data points. This is because ElastiCache Auto Scaling always acts conservatively by rounding up or down when it determines how much capacity to add or remove. This prevents it from adding insufficient capacity or removing too much capacity.

  • To ensure application availability, the service scales out proportionally to the metric as fast as it can, but scales in more conservatively.

  • You can have multiple target tracking scaling policies for an ElastiCache for Valkey and Redis OSS cluster, provided that each of them uses a different metric. The intention of ElastiCache Auto Scaling is to always prioritize availability, so its behavior differs depending on whether the target tracking policies are ready for scale out or scale in. It will scale out the service if any of the target tracking policies are ready for scale out, but will scale in only if all of the target tracking policies (with the scale-in portion enabled) are ready to scale in.

  • Do not edit or delete the CloudWatch alarms that ElastiCache Auto Scaling manages for a target tracking scaling policy. ElastiCache Auto Scaling deletes the alarms automatically when you delete the scaling policy.

  • ElastiCache Auto Scaling doesn't prevent you from manually modifying cluster shards. These manual adjustments don't affect any existing CloudWatch alarms that are attached to the scaling policy but can impact metrics that may trigger these CloudWatch alarms.

  • These CloudWatch alarms managed by Auto Scaling are defined over the AVG metric across all the shards in the cluster. So, having hot shards can result in either scenario of:

    • scaling when not required due to load on a few hot shards triggering a CloudWatch alarm

    • not scaling when required due to aggregated AVG across all shards affecting alarm not to breach.

  • ElastiCache default limits on Nodes per cluster still applies. So, when opting for Auto Scaling and if you expect maximum nodes to be more than default limit, request a limit increase at AWS Service Limits and choose the limit type Nodes per cluster per instance type.

  • Ensure that you have enough ENIs (Elastic Network Interfaces) available in your VPC, which are required during scale-out. For more information, see Elastic network interfaces.

  • If there is not enough capacity available from EC2, ElastiCache Auto Scaling would not scale and be delayed til the capacity is available.

  • ElastiCache for Redis OSS Auto Scaling during scale-in will not remove shards with slots having an item size larger than 256 MB post-serialization.

  • During scale-in it will not remove shards if insufficient memory available on resultant shard configuration.