AWS has previously released an Application Auto Scaling model that automates capacity management for DynamoDB tables. One feature of this model is target tracking. Each Auto Scaling policy with target tracking requires that you choose a target value for a CloudWatch metric. Auto Scaling will then drive the metric towards this specified target and simultaneously adjust CloudWatch alarms.
This is easier and more direct as opposed to having to set up ranges and thresholds with the old step-scaling policy. But you still have the option to use target tracking with step scaling to create an advanced scaling policy.
AWS just announced that target tracking has been added to EC2 Auto Scaling. Now you may generate scaling policies that are driven by a metric you choose: application load balancer request counts, CPU load, network traffic, and so on.
Please note that these metrics mentioned share an essential property—namely, any additional EC2 instances will drive the metric down, and fewer instances will drive them up.
Please note the following when creating your target tracking policies:
- You can track several targets in a single Auto Scaling Group so long as they each reference a separate metric. Scaling will always choose the policy the drives the highest capacity.
- Scaling will not occur with metrics that do not have enough data.
- Auto Scaling will compensate for abrupt fluctuations in the metrics and will try to minimize related fluctuations in capacity.
- You may use target tracking for a custom metric using the Auto Scaling API or the AWS CLI.
- Normally you will want to scale on metrics that get published with 1-minute cycles. This is called detailed monitoring. Using 5-minutes will mean a slower response time.
- This service comes free. You can start using it now.
Contact us on PolarSeven now if you would like help bringing your enterprise into the AWS cloud.