Managing throughput capacity automatically with DynamoDB auto scaling

Many database workloads are cyclical in nature, while others are difficult to predict in advance. For one example, consider a social networking app where most of the users are active during daytime hours. The database must be able to handle the daytime activity, but there's no need for the same levels of throughput at night. For another example, consider a new mobile gaming app that is experiencing unexpectedly rapid adoption. If the game becomes too popular it could exceed the available database resources, resulting in slow performance and unhappy customers. These kinds of workloads often require manual intervention to scale database resources up or down in response to varying usage levels.

Amazon DynamoDB auto scaling uses the AWS Application Auto Scaling service to dynamically adjust provisioned throughput capacity on your behalf, in response to actual traffic patterns. This enables a table or a global secondary index (GSI) to increase its provisioned read and write capacity to handle sudden increases in traffic, without throttling. When the workload decreases, Application Auto Scaling decreases the throughput so that you don't pay for unused provisioned capacity.

With Application Auto Scaling, you create a scaling policy for a table or a global secondary index. The scaling policy specifies whether you want to scale read capacity or write capacity (or both), and the minimum and maximum provisioned capacity unit settings for the table or index.

The scaling policy also contains a target utilization—the percentage of consumed provisioned throughput at a point in time. Application Auto Scaling uses a target tracking algorithm to adjust the provisioned throughput of the table (or index) upward or downward in response to actual workloads, so that the actual capacity utilization remains at or near your target utilization.

DynamoDB outputs consumed provisioned throughput for one-minute periods. Auto scaling triggers when your consumed capacity breaches the configured target utilization for two consecutive minutes. CloudWatch alarms might have a short delay of up to a few minutes before triggering auto scaling. This delay ensures accurate CloudWatch metric evaluation. If the consumed throughput spikes are more than a minute apart, auto scaling might not trigger. Similarly, a scale down event can occur when 15 consecutive data points are lower than the target utilization. In either case, after auto scaling triggers, the UpdateTable API is invoked. It then takes several minutes to update the provisioned capacity for the table or index. During this period, any requests that exceed the previous provisioned capacity of the tables are throttled.

You can set the auto scaling target utilization values between 20 and 90 percent for your read and write capacity.

How DynamoDB auto scaling works

The following diagram provides a high-level overview of how DynamoDB auto scaling manages throughput capacity for a table.

The following steps summarize the auto scaling process as shown in the previous diagram:

To understand how DynamoDB auto scaling works, suppose that you have a table named ProductCatalog. The table is bulk-loaded with data infrequently, so it doesn't incur very much write activity. However, it does experience a high degree of read activity, which varies over time. By monitoring the Amazon CloudWatch metrics for ProductCatalog, you determine that the table requires 1,200 read capacity units (to avoid DynamoDB throttling read requests when activity is at its peak). You also determine that ProductCatalog requires 150 read capacity units at a minimum, when read traffic is at its lowest point. For more information about preventing throttling, see Throttling issues for DynamoDB.

Within the range of 150 to 1,200 read capacity units, you decide that a target utilization of 70 percent would be appropriate for the ProductCatalog table. Target utilization is the ratio of consumed capacity units to provisioned capacity units, expressed as a percentage. Application Auto Scaling uses its target tracking algorithm to ensure that the provisioned read capacity of ProductCatalog is adjusted as required so that utilization remains at or near 70 percent.

To enable DynamoDB auto scaling for the ProductCatalog table, you create a scaling policy. This policy specifies the following:

When you create a scaling policy, Application Auto Scaling creates a pair of Amazon CloudWatch alarms on your behalf. Each pair represents the upper and lower boundaries for your provisioned throughput settings. These CloudWatch alarms are triggered when the table's actual utilization deviates from your target utilization for a sustained period of time.

When one of the CloudWatch alarms is triggered, Amazon SNS sends you a notification (if you have enabled it). The CloudWatch alarm then invokes Application Auto Scaling, which in turn notifies DynamoDB to adjust the ProductCatalog table's provisioned capacity upward or downward as appropriate.

During a scaling event, AWS Config is charged per configuration item recorded. When a scaling event occurs, four CloudWatch alarms are created for each read and write auto-scaling event: ProvisionedCapacity alarms: ProvisionedCapacityLow, ProvisionedCapacityHigh and ConsumedCapacity alarms: AlarmHigh, AlarmLow. This results in a total of eight alarms. Therefore, AWS Config records eight configuration items for every scaling event.

Usage notes

Before you begin using DynamoDB auto scaling, you should be aware of the following: