Skip to content
16 changes: 12 additions & 4 deletions tidb-cloud/architecture-concepts.md
Original file line number Diff line number Diff line change
Expand Up @@ -223,9 +223,10 @@ One RCU represents a sustained capacity of RUs per second. For example, a baseli

### RCU auto-scaling

When configuring your {{{ .premium }}} instance, you specify the maximum number of RCUs (`RCU_max`) required for your workload. TiDB Cloud automatically scales capacity within the range of `0.25 * RCU_max` to `RCU_max`.
TiDB Cloud automatically scales the capacity of your {{{ .premium }}} instance based on workload demand. The scaling behavior depends on the capacity mode:

For example, if you set the maximum capacity to 20,000 RCUs, TiDB Cloud dynamically scales the capacity between 5,000 and 20,000 RCUs based on real-time demand. This scaling is automatic and instantaneous, enabling you to consume up to the maximum number of RCUs at any time without manual intervention or delay.
- In **Elastic Mode**, you configure a Baseline RCU instead of a maximum RCU. Baseline RCU is the initial capacity for autoscaling, and TiDB Cloud scales capacity beyond this baseline as workload demand increases.
- In **Limited Mode**, you configure the maximum number of RCUs (`RCU_max`) required for your workload. TiDB Cloud automatically scales capacity within the range of `0.25 * RCU_max` to `RCU_max`. For example, if you set the maximum capacity to 20,000 RCUs, TiDB Cloud dynamically scales capacity between 5,000 and 20,000 RCUs based on real-time demand.

### RCU billing

Expand All @@ -237,9 +238,16 @@ TiDB Cloud calculates your usage every minute. It measures the total number of R

#### Minimum usage requirement

To maintain baseline capacity and ensure that resources are always available for your instance, TiDB Cloud automatically sets a minimum billing RCU based on your maximum RCU setting. This value defines the baseline reserved capacity for your instance.
To maintain baseline capacity and ensure that resources are always available for your instance, TiDB Cloud applies a minimum billing RCU. This value defines the reserved capacity for your instance and is the minimum number of RCUs you are billed for.

If your actual consumption in a given minute is below this threshold, billing defaults to the minimum billing RCU. This mechanism ensures that your instance can immediately handle sudden traffic spikes up to your specified maximum, without performance degradation or delays.
If your actual consumption in a given minute is below this threshold, billing defaults to the minimum billing RCU. This mechanism provides baseline capacity to handle incoming traffic while allowing the instance to scale out as demand increases.

How the minimum billing RCU is determined depends on the capacity mode of your instance:

- In **Elastic Mode**, you set a Baseline RCU as the minimum billing RCU for your instance. Baseline RCU is also the initial capacity for autoscaling, that is, the capacity your instance has before it scales out to handle increased workload. If your actual consumption in a given minute is lower than the Baseline RCU, billing is calculated using the Baseline RCU.
- In **Limited Mode**, you set the maximum capacity that your instance can scale up to. TiDB Cloud automatically sets the minimum billing RCU to the greater of 25% of your configured maximum RCU and 5,000 RCUs.

Baseline RCU does not change the minimum monthly plan charge for your {{{ .premium }}} instance.

### Request Unit (RU)

Expand Down
5 changes: 4 additions & 1 deletion tidb-cloud/backup-and-restore-serverless.md
Original file line number Diff line number Diff line change
Expand Up @@ -122,7 +122,10 @@ To restore your data to a new {{{ .starter }}} or Essential instance, take the f
5. Update the capacity as needed.

- For a {{{ .starter }}} instance, if you need more resources than the [free quota](/tidb-cloud/select-cluster-tier.md#usage-quota), set a monthly spending limit.
- For a {{{ .essential }}} instance, set the minimum RCU and maximum RCU, and then configure advanced settings as needed.
- For a new {{{ .essential }}} instance, configure the following capacity settings based on the TiDB kernel version, and then configure advanced settings as needed.

- For a new {{{ .essential }}} instance running on CLOUD.202603.x or later, **Elastic Mode** is selected by default and uses a fixed **Baseline RCU** of 2,000 RCUs. If you need to limit capacity, switch to **Limited Mode** and set the maximum RCU; the minimum billing RCU is calculated automatically.
- For a new {{{ .essential }}} instance running on earlier TiDB versions, set the minimum RCU and maximum RCU.

6. Click **Restore** to begin the restore process.

Expand Down
8 changes: 5 additions & 3 deletions tidb-cloud/create-tidb-cluster-serverless.md
Original file line number Diff line number Diff line change
Expand Up @@ -60,9 +60,11 @@ If you are in the `Organization Owner` or the `Project Owner` role, you can crea

- **Essential** plan:

- You must specify both a minimum and maximum number of Request Capacity Units (RCUs) for your {{{ .essential }}} instance.
- For {{{ .essential }}} instances running on CLOUD.202603.x or later, **Elastic Mode** is selected by default. In Elastic Mode, you do not set a maximum number of RCUs. Instead, your instance uses a fixed **Baseline RCU** of 2,000 RCUs, which is the minimum number of Request Capacity Units (RCUs) you are billed for and the initial capacity for autoscaling.

- RCUs represent the compute resources provisioned for your workload. TiDB Cloud automatically scales your {{{ .essential }}} instance within this range based on demand.
- To limit the capacity, select **Limited Mode**, and then specify the maximum number of RCUs that your instance can scale up to. {{{ .essential }}} instances on earlier versions do not support Elastic Mode; for these instances, specify both the minimum and maximum number of RCUs.

- RCUs represent the compute resources provisioned for your workload. TiDB Cloud automatically scales your {{{ .essential }}} instance based on demand. If your workload consistently requires more than 100,000 RCUs, consider upgrading to {{{ .premium }}} to avoid potential bottlenecks.

7. Click **Create**.

Expand All @@ -76,4 +78,4 @@ If you are in the `Organization Owner` or the `Project Owner` role, you can crea
>
> If you do not set a password, you cannot connect to the {{{ .starter }}} or Essential instance.

- To manage your instance using the TiDB Cloud CLI, see [TiDB Cloud CLI Reference (PREVIEW)](/tidb-cloud/cli-reference.md).
- To manage your instance using the TiDB Cloud CLI, see [TiDB Cloud CLI Reference (PREVIEW)](/tidb-cloud/cli-reference.md).
2 changes: 2 additions & 0 deletions tidb-cloud/monitor-built-in-alerting.md
Original file line number Diff line number Diff line change
Expand Up @@ -131,6 +131,7 @@ TiDB Cloud provides different alert rules for each [TiDB Cloud plan](/tidb-cloud
| Condition | Recommended Action |
|:--- |:--- |
| Request units per second (RU/s) exceed 80% of the maximum RCU | <ol><li>Review RU metrics to determine whether the increase is gradual or a sudden spike.</li><li>If the increase is gradual, check whether query duration has increased. If so, the current maximum RCU might be insufficient.</li><li>Scale capacity by manually increasing the maximum RCU in the TiDB Cloud console.</li></ol><br/>If you cannot resolve the issue, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md).|
| RCU consumption (Request Unit/s) exceeds 100,000 RCUs | <ol><li>Check your traffic pattern in the RU metrics to determine whether the increase is gradual or a sharp spike.</li><li>In Elastic Mode, TiDB Cloud automatically scales backend capacity with your workload. If this usage level is expected, adjust the alert threshold to match your workload. Note that sustained high RU/s may increase your costs.</li><li>If this usage level is unexpected, use the Top RU feature to identify and optimize the queries consuming the most RUs.</li></ol><br/>If you cannot resolve the issue, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md).|
| QPS drops by 80% | <ol><li>Check whether the drop is caused by increasing query latency.</li><li>Verify that your application is operating normally. If the drop is intentional, ignore this alert. If the drop is unintentional and you cannot identify the root cause, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md) immediately.</li></ol>|
| Query P99 latency exceeds 200 ms | <ol><li>Investigate slow queries: go to the Slow Query page and filter by a recent time range to identify newly introduced or slower-running queries.</li><li>Review recent changes, such as application deployments, schema changes, or data import jobs, that might have affected traffic patterns.</li></ol><br/>If you cannot identify the root cause, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md) immediately.|
| Query P95 latency exceeds 200 ms | <ol><li>Investigate slow queries: go to the Slow Query page and filter by a recent time range to identify newly introduced or slower-running queries.</li><li>Review recent changes, such as application deployments, schema changes, or data import jobs, that might have affected traffic patterns.</li></ol><br/>If you cannot identify the root cause, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md) immediately.|
Expand All @@ -156,6 +157,7 @@ TiDB Cloud provides different alert rules for each [TiDB Cloud plan](/tidb-cloud
| Condition | Recommended Action |
|:--- |:--- |
| Request units per second (RU/s) exceed 80% of the maximum RCU | <ol><li>Review RU metrics to determine whether the increase is gradual or a sudden spike.</li><li>If the increase is gradual, check whether query duration has increased. If so, the current maximum RCU might be insufficient.</li><li>Scale capacity by manually increasing the maximum RCU in the TiDB Cloud console.</li></ol><br/>If you cannot resolve the issue, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md).|
| RCU consumption (Request Unit/s) exceeds 100,000 RCUs | <ol><li>Check your traffic pattern in the RU metrics to determine whether the increase is gradual or a sharp spike.</li><li>In Elastic Mode, TiDB Cloud automatically scales backend capacity with your workload. If this usage level is expected, consider upgrading to TiDB Cloud Premium if consumption remains above 100,000 RCUs to avoid potential performance bottlenecks. Note that sustained high RU/s may increase your costs.</li><li>If this usage level is unexpected, use the Top RU feature to identify and optimize the queries consuming the most RUs.</li></ol><br/>If you cannot resolve the issue, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md).|
| QPS drops by 80% | <ol><li>Check whether the drop is caused by increasing query latency.</li><li>Verify that your application is operating normally. If the drop is intentional, ignore this alert. If the drop is unintentional and you cannot identify the root cause, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md) immediately.</li></ol>|
| Query P99 latency exceeds 200 ms | <ol><li>Investigate slow queries: go to the Slow Query page and filter by a recent time range to identify newly introduced or slower-running queries.</li><li>Review recent changes, such as application deployments, schema changes, or data import jobs, that might have affected traffic patterns.</li></ol><br/>If you cannot identify the root cause, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md) immediately.|
| Query P95 latency exceeds 200 ms | <ol><li>Investigate slow queries: go to the Slow Query page and filter by a recent time range to identify newly introduced or slower-running queries.</li><li>Review recent changes, such as application deployments, schema changes, or data import jobs, that might have affected traffic patterns.</li></ol><br/>If you cannot identify the root cause, contact [TiDB Cloud Support](/tidb-cloud/tidb-cloud-support.md) immediately.|
Expand Down
10 changes: 7 additions & 3 deletions tidb-cloud/premium/create-tidb-instance-premium.md
Original file line number Diff line number Diff line change
Expand Up @@ -42,12 +42,16 @@ If you have the `Organization Owner` role, you can create a {{{ .premium }}} ins
2. On the **Create Resource** page, select **Premium** as your plan.
3. Enter a name for your {{{ .premium }}} instance, and then choose a cloud provider and a region where you want to host your instance.
4. (Optional) To group this {{{ .premium }}} instance in a project for management, click **Group Your Instance in a Project**, and then select the target project for the instance. If there is no project in your organization, you can create one by clicking **Create a Project**.
5. In the **Capacity** area, set the maximum number of Request Capacity Units (RCUs) for your instance. RCUs represent the compute resources provisioned for your workload. TiDB Cloud automatically scales your instance within this range based on demand.
5. In the **Capacity** area, configure the capacity of your instance. **Elastic Mode** is selected by default.

- **Elastic Mode**: set a **Baseline RCU** for your instance (rather than a maximum RCU). Baseline RCU is the minimum number of Request Capacity Units (RCUs) you are billed for and the initial capacity for autoscaling. TiDB Cloud automatically allocates backend resources based on your workload and scales out from the Baseline RCU when demand increases. If your actual consumption in a given minute is lower than the Baseline RCU, billing is calculated using the Baseline RCU.

- **Limited Mode**: set the maximum capacity that your instance can scale up to. RCUs represent the compute resources provisioned for your workload. TiDB Cloud automatically scales your instance within the range of `0.25 * RCU_max` to `RCU_max` based on demand.

> **Note:**
>
> - The maximum RCU value must be set in increments of 100.
> - You will be billed for a minimum number of RCUs, known as the **Minimum Billing RCU**. This value is the greater of 25% of your configured maximum RCU and 5,000 RCUs. The minimum billing RCU ensures that dedicated resources are reserved for peak performance. You are charged this minimum amount even if your actual usage is lower.
> - In **Limited Mode**, the maximum RCU value must be set in increments of 100.
> - In **Limited Mode**, you are billed for a minimum number of RCUs, known as the **Minimum Billing RCU**. This value is the greater of 25% of your configured maximum RCU and 5,000 RCUs. The minimum billing RCU ensures that dedicated resources are reserved to maintain baseline capacity. You are charged this minimum amount even if your actual usage is lower.

6. For {{{ .premium }}} instances, only regional high availability is enabled, and it is not configurable. For more information, see [High Availability](/tidb-cloud/serverless-high-availability.md).

Expand Down
4 changes: 2 additions & 2 deletions tidb-cloud/tidb-cloud-glossary.md
Original file line number Diff line number Diff line change
Expand Up @@ -168,8 +168,8 @@ TiDB Cloud measures the capacity of [changefeeds](/tidb-cloud/changefeed-overvie

For {{{ .essential }}} and {{{ .premium }}}, a Request Capacity Unit (RCU) is a unit of measure used to represent the provisioned compute capacity for your {{{ .essential }}} or {{{ .premium }}} instance. One RCU provides a fixed amount of compute resources that can process a certain number of RUs per second. The number of RCUs you provision determines the baseline performance and throughput capacity of your instance. However, the way RCUs are managed differs between {{{ .essential }}} and {{{ .premium }}}:

- {{{ .essential }}} automatically provisions RCUs based on your workload. As your QPS increases, TiDB Cloud dynamically scales up the provisioned RCUs to maintain performance. For more information, see [{{{ .essential }}} Pricing Details](https://www.pingcap.com/tidb-cloud-essential-pricing-details/).
- {{{ .premium }}} lets you specify the maximum number of RCUs (`RCU_max`) for your workload. TiDB Cloud automatically scales capacity within the range of `0.25 * RCU_max` to `RCU_max` based on real-time demand. For more information, see [Request units and capacity in {{{ .premium }}}](https://docs.pingcap.com/tidbcloud/architecture-concepts/?plan=premium#request-units-and-capacity-in-premium).
- {{{ .essential }}} instances running on CLOUD.202603.x or later support **Elastic Mode** and **Limited Mode**. In Elastic Mode, the instance uses a fixed Baseline RCU of 2,000 RCUs and scales capacity based on workload demand. In Limited Mode, you configure the maximum RCU. For instances on earlier versions, you configure both the minimum and maximum RCUs. For more information, see [{{{ .essential }}} Pricing Details](https://www.pingcap.com/tidb-cloud-essential-pricing-details/).
- {{{ .premium }}} supports two capacity modes. In **Elastic Mode**, you configure a Baseline RCU, and TiDB Cloud scales capacity beyond this baseline as workload demand increases. In **Limited Mode**, you configure the maximum number of RCUs (`RCU_max`), and TiDB Cloud automatically scales capacity within the range of `0.25 * RCU_max` to `RCU_max`. For more information, see [Request units and capacity in {{{ .premium }}}](https://docs.pingcap.com/tidbcloud/architecture-concepts/?plan=premium#request-units-and-capacity-in-premium).

### Request Unit (RU)

Expand Down
Loading