This page explains how autoscaling works and helps you determine if autoscaling is right for your use case. Before you read this page, you should be familiar with the Overview of Bigtable and Instances, clusters, and nodes.
In Cloud Bigtable, instances are containers for clusters, which are location-specific resources that handle requests. Each cluster has one or more nodes, which are compute resources used to manage your data. When you create a cluster in an instance, you choose either manual node allocation or autoscaling.
With manual node allocation, the number of nodes in the cluster remains constant until you change it. When autoscaling is enabled, Bigtable continuously monitors the cluster and automatically adjusts the number of nodes in the cluster when necessary. Autoscaling works on both HDD and SSD clusters, in all Bigtable regions.
You can configure autoscaling in the Google Cloud console, using
gcloud, or using
the Cloud Bigtable client library for Java.
When to use autoscaling
The benefits of autoscaling include the following:
- Costs - Autoscaling can help you optimize costs because Bigtable reduces the number of nodes in your cluster whenever possible. This can help you avoid over-provisioning.
- Performance - Autoscaling lets Bigtable automatically add nodes to a cluster when a workload changes or there is an increase in data storage requirements. This helps maintain workload performance objectives by ensuring that the cluster has enough nodes to meet the target CPU utilization and storage requirements.
- Automation - Autoscaling reduces management complexity. You don't need to monitor and scale the cluster size manually or write an application to do these tasks, because the Bigtable service handles them for you.
Autoscaling is often the best choice for the following situations:
- Steady diurnal traffic patterns such as those generated by online shopping services
- New applications expecting organic growth
- Workloads that are new to Bigtable
Autoscaling might not work well for the following workload types, because even though Bigtable quickly adds nodes when traffic increases, it can take time to balance the additional nodes.
- Bursty traffic
- Sudden batch workloads
If your spikes in usage are predictable or regularly scheduled, you might be able to use autoscaling and adjust the settings before the planned bursts. See Delay while nodes rebalance for details.
How autoscaling works
Autoscaling is the process of automatically scaling, or changing the size of, a cluster by adding or removing nodes. When you enable autoscaling, Bigtable automatically adjusts the size of your cluster for you. When your cluster's workload or storage needs fluctuate, Bigtable either scales up, adding nodes to the cluster, or it scales down, removing nodes from the cluster.
Bigtable autoscaling determines the number of nodes required, based on the following dimensions:
- CPU utilization target
- Storage utilization target
- Minimum number of nodes
- Maximum number of nodes
Each scaling dimension generates a recommended node count, and Bigtable automatically uses the highest one. This means, for example, that if your cluster needs 10 nodes to meet your storage utilization target but 12 to meet your CPU utilization target, Bigtable scales the cluster to 12 nodes.
As the number of nodes changes, Bigtable continuously optimizes the storage, rebalancing data across the nodes, to ensure that traffic is spread evenly and no node is overloaded.
After a cluster is scaled up, Bigtable automatically rebalances the nodes in your cluster for optimal performance. All requests continue to reach the cluster while scaling and rebalancing are in progress. See Scaling limitations for more information.
If a cluster has scaled up to its maximum number of nodes and the CPU utilization target is exceeded, requests might have high latency or fail. If a cluster has scaled up to its maximum number of nodes and the storage utilization limit is exceeded, write requests will fail. See Storage per node for more details on storage limits.
When a cluster is scaled down, nodes are removed at a slower rate than when scaling up, to prevent any impact on latency. See scaling limitations for more details.
When you create or edit a cluster and choose autoscaling, you define the values for CPU utilization target, min nodes, and max nodes. You can either configure the storage utilization target or leave it at the default, which is 50% (2.5 TB for SSD and 8 TB for HDD).
|CPU utilization target||
A percentage of the cluster's CPU capacity. Can be from 10% to 80%. When a cluster's CPU utilization exceeds the target that you have set, Bigtable immediately adds nodes to the cluster. When CPU utilization is substantially lower than the target, Bigtable removes nodes.
|Minimum number of nodes||
The lowest number of nodes that Bigtable will scale the cluster down to. This value must be greater than zero and can't be lower than 10% of the value you set for the maximum number of nodes. For example, if the maximum number of nodes is 40, the minimum number of nodes must be at least 4. The 10% requirement is a hard limit.
|Maximum number of nodes||
The highest number of nodes that you want to let the cluster scale up to. This value must be greater than zero and equal to or greater than the minimum number of nodes. The value can't be more than 10 times the number that you choose for the minimum number of nodes. This 10x requirement is a hard limit.
|Storage utilization target||
The maximum number of terabytes per node that you can store before Bigtable scales up. This target ensures that you always have enough nodes to handle fluctuations in the amount of data that you store.
This section describes how to choose your autoscaling parameters. After you set your initial values, monitor your cluster and adjust the numbers if necessary.
Determine the CPU utilization target
Base the CPU utilization target on your unique workload. The optimal target for your cluster depends on the latency and throughput requirements of your workload. See CPU usage for recommendations, and see Trade-off between high throughput and low latency to understand the reasons behind the recommendations.
In general, if you observe unacceptably high latency, you should lower the CPU utilization target.
Determine the storage utilization target
If your application is latency-sensitive, keep storage utilization below 60%. If your application is not latency-sensitive, you can choose a storage utilization target of 70% or more. For details, see Plan your Bigtable capacity.
For autoscaling, storage utilization is expressed as the number of bytes of storage per node rather than as a percentage. The storage utilization target is specified per node but is applied to the entire cluster. The capacity limits for nodes are 5 TB per node for SSD storage and 16 TB per node for HDD storage.
The following table shows target amounts for typical storage utilization target percentages. The Google Cloud console accepts the value in TB per node, and the gcloud CLI, API, and Bigtable client libraries accept an integer value in GiB per node.
|80%||4 TB or 4,096 GiB||12.8 TB or 13,107 GiB|
|70%||3.5 TB or 3,584 GiB||11.2 TB or 11,468 GiB|
|60%||3 TB or 3,072 GiB||9.6 TB or 9,830 GiB|
|50%||2.5 TB or 2,560 GiB||8 TB or 8,192 GiB|
Determine the maximum number of nodes
The value that you choose as the maximum number of nodes should be the number of nodes that the cluster needs to handle your workload's heaviest traffic, even if you don't expect to reach that volume most of the time. Bigtable never scales up to more nodes than it needs. You can also think of this number as the highest number of nodes that you are willing to pay for. See Autoscaling parameters for details on accepted values.
The maximum number needs to allow for both the CPU utilization target set by you and the storage utilization target set by Bigtable.
If you're changing a cluster from manual allocation to autoscaling, find the highest number of nodes that the cluster has had over the last month or so. Your autoscaling maximum should be at least that number.
If you are enabling autoscaling for a new cluster in an existing instance, use metrics from other clusters in the instance as your guidepost.
If you have a new workload and you're not sure how it's going to grow, you can estimate the number of nodes that you need to meet the built-in storage utilization target and then adjust the number later.
To arrive at this number, estimate the amount of data that you plan to store on the cluster, and then divide that number by the storage utilization target for the storage type that you use.
For example, if you store 10 TB on an SSD cluster, you can divide 10 TB by 2.5 TB, which is the storage utilization target set by default for SSD clusters that use autoscaling. The result is 4, which means that 4 is the number of nodes that can handle that amount of data, and the maximum should be some number higher than that.
Using the same formula, the following examples show the number of nodes you might need for some sample storage amounts:
|SSD storage per cluster||Smallest maximum number of nodes|
After your cluster is up and running with autoscaling enabled, monitor the
cluster and make sure that the value you choose for maximum number
of nodes is at least as high as the
recommended number of nodes for CPU target
recommended number of nodes for storage target.
Determine the minimum number of nodes
You can set the minimum as low as 1 to ensure that your Bigtable can scale down to the smallest, most cost-efficient size if possible. The cluster never becomes too small because Bigtable automatically prevents the node count from dropping below the minimum needed to maintain the CPU and storage utilization targets. See Autoscaling parameters for details on accepted values.
However, in many cases you will want to set this value to more than one. Choose a higher number or raise the minimum number of nodes for the following situations:
- You have an upcoming event, such as Cyber Monday, when you expect your traffic to temporarily increase, and you want to make sure you have enough capacity.
- Your application sends spiky traffic. When new nodes are added, Bigtable automatically rebalances onto the new nodes. Because this process can take several minutes, you are often better off taking a conservative approach and choosing a higher minimum so that your cluster can seamlessly accommodate the spikes.
- You increase the maximum number of nodes. The minimum must always be ten percent or less of the maximum number of nodes. For example, if you set the maximum to 30, you must set the minimum to at least 3.
Fine-tune your settings
Keep an eye on your node usage and adjust your settings if necessary, especially after you first enable autoscaling.
Account for replication
In an instance that uses replication, each cluster's autoscaling settings and activity are completely independent of those for the other clusters in the instance. You must configure the scaling mode for each cluster in an instance.
In general, for multi-cluster instances, you should enable autoscaling for every cluster in the instance. Your autoscaling configuration will often be the same for every cluster in the instance, but might be different depending on the use case, workload, and performance requirements for each cluster.
Because clusters in a replicated instance do some additional work to manage replication, you should choose a higher maximum number of nodes for them than you would for a single-cluster instance. To learn more, see Replication and performance.
To configure autoscaling, you need to be a principal in a role that has
for the cluster and instance that you are configuring.
Bigtable provides several metrics to help you understand how Bigtable autoscaling is working as it scales up and down to meet workload requirements. The metrics also can help you gauge whether your settings are optimal to meet your business's workload and cost requirements. For example, if you observe that the node count for a cluster is often close to the maximum number of nodes, you might consider raising the maximum. To learn more about monitoring your Bigtable resources, see Monitoring an instance.
Bigtable emits a system event audit log each time it scales a cluster. The log entry is similar to the following:
Grew from 9 to 10 nodes to maintain CPU utilization at 60%.
You can view autoscaling system event logs on the Bigtable cluster overview page in the Google Cloud console. You can also view them using Logs Explorer:
Navigate to the Logs Explorer:
Select the appropriate Cloud project.
In the Query field, enter the following:
resource.type="audited_resource" resource.labels.service="bigtableadmin.googleapis.com" resource.labels.method="AutoscaleCluster"
Click Run query.
The Query results pane displays the logs for the last hour.
To learn more about viewing logs, see Cloud Logging.