A compute profile is a scaling policy for compute clusters in the same compute group that are the same size. These compute clusters have the same compute map.
A compute group can have one or more compute profiles (for example, a compute profile for large compute clusters and a compute profile for small compute clusters). A compute profile controls the size of the compute group using the INSTANCE clause, which specifies a compute map. Map entries range from TD_COMPUTE_XSMALL (1 node) to TD_COMPUTE_2XL (32 nodes). See compute_map in CREATE COMPUTE PROFILE Syntax Elements.
Because you pay for active compute profiles, Teradata recommends having only one active compute profile at a time, except for a brief overlap when one is taking over from another (the cooldown period). (To see how your organization consumes compute and storage resources, see Review Consumption Usage.)
Switching between compute profiles can be accomplished using a schedule (for examples, see Example: Schedule a Small Compute Cluster for 5:00 PM - 8:00 AM and a Large Compute Cluster for 8:00 AM - 5:00 PM and Example: Schedule a Compute Cluster for Monday-Friday 8:00 AM - 5:00 PM). Queries running on the current compute profile complete only if they can do so before its cooldown period starts, when new queries run on the new compute profile.
Specifying Automatic Scaling
Automatic scaling manages compute clusters based on workload. By providing only the necessary compute clusters, automatic scaling reduces cost.
To create a compute profile that specifies automatic scaling, specify the scaling range with MIN_COMPUTE_COUNT and MAX_COMPUTE_COUNT. For descriptions of these arguments, see CREATE COMPUTE PROFILE Syntax Elements.
How Automatic Scaling Works
Automatic scaling monitors the workload of a compute group, specifically the following values:
| Monitored Value | Description |
|---|---|
| CPU | Average percentage of active compute clusters in the compute group. |
| Memory | Estimated available memory for active and new queries in the compute group. |
| Count | Number of concurrent queries that can run on a compute cluster in the compute group. |
Based on the monitored values, automatic scaling scales out or scales in.
| Term | Description | When It Happens |
|---|---|---|
| Scaling out | Activates more compute clusters, up to MAX_COMPUTE_COUNT. | When any of the following is true for two minutes:
|
| Scaling in | Hibernates some active compute clusters, down to MIN_COMPUTE_COUNT. Existing work continues until cooldown period expires and any unfinished work has restarted on remaining active clusters. |
When CPU, Memory, and Count for two or more active compute clusters is below capacity to hibernate compute clusters for two minutes. |
After scaling out or scaling in (which takes minutes), new work is distributed between the active compute clusters.
Compute Profile Status in VantageCloud Lake Console
You can view the Status of a profile on the Compute groups screen for each profile. To view the Status description, select next to the compute profile then select Edit.
The following table shows the Status and the description for each message that appears at the top of the Edit profile dialog.
| Status | Description |
|---|---|
| Deploying | Provisioning your resources |
| Deploying - awaiting resources | Your cloud service provider does not have sufficient resources to complete the deployment. We'll keep trying. |
| Partially active | You can use the current resources while the system completes the provisioning process. |
| Partially active - waiting for resources | You can use the current resources while the system completes the provisioning process. However, your cloud service provider does not have sufficient resources to complete the deployment. We'll keep trying. |
| Active | You can use the requested resources. Additional resources may become available, based on your autoscaling policy and workload needs. |
| Active - Scaling out | Provisioning additional resources based on your autoscaling policy and workload needs. |
| Active - waiting for resources | Provisioning additional resources based on your autoscaling policy and workload needs. However, your cloud service provider does not have sufficient resources. We'll keep trying. |
| Active - Scaling in | Shutting down unneeded resources based on your autoscaling policy and workload needs. |
| Suspended | Shutting down all resources based on an immediate request or schedule. |
| Suspending - cooling down | Shutting down all resources based on an immediate request or schedule. Currently running queries will continue to run until completed or the cooldown period ends, whichever comes first. |
| Suspending | Shutting down all resources based on an immediate request or schedule. No more queries can be run on this resource. |
| Resuming | Resuming resources based on an immediate request or schedule. |
| Resuming - waiting for resources | Resuming resources based on an immediate request or schedule. However, your cloud service provider does not have sufficient resources. We'll keep trying. |
| Deploying - Recovering from failure | Re-provisioning resources to recover from failure due to issues related to cloud service provider resources or software. |