Follow these steps to create a compute group and its compute profile. The compute profile defines the number of clusters, the degree of autoscale, and the schedule window for the compute group when the clusters run. A compute group can have multiple compute profiles for different times of day or different workloads.
After the compute group is created, you can set the initial users. They can be group admins or group users. A group admin can manually suspend and resume compute groups even when a schedule is defined. Group users can run their workloads on this compute group. Users can belong to more than one group and choose which group to run which workload.
For more information, see:
- Select the environment, then Compute Groups.
- Enter the dbc name and password.
- Select .
- Enter a Name and Description for the new compute group.
- Select the Group type and select Create.
- Standard - Suitable for a wide variety of queries and in-database analytics.
- Analytic - Suitable for memory-intensive analytic queries to run in open analytics framework.
- Analytic GPU - Configured with NVIDIA GPU accelerated containers that are designed for model inference jobs using deep learning and large language models.
- Select the new group you just created, then under Profiles, select Create.
- Fill out Create profile and select Create.
Item Description Name Name for this profile Instance size - XSmall
- Small
- Medium
- Large
- XLarge
- XXLarge
Availability Always
On schedule- Day - Set the Start time and End time in UTC
- Week - Select day of week and Start time and End time in UTC
- Month - Select day of month and Start time and End tim in UTC
Suspended- If you suspend when you create the profile, the profile activates at the start time in the schedule, saving you consumption units.
- If the profile is already active when you suspend, schedules are skipped until resumed.
Auto-scale If on, automatically activate and suspend nodes to meet variable demand. - If off, select the Number of clusters.
If on, set Minimum and Maximum number of clusters.
Cool-down time Select the number of minutes to allow queries to run after the End time. You cannot start new queries during the cool-down time.