The advantage of Open Analytics Framework is that clusters within a compute group can be instantiated as needed and scaled on demand or on schedule. Scaling is regulated by specifying one or more analytic or analytic GPU compute profiles. The compute profile is associated with the compute group to use, determines availability schedule, auto-scaling, and cool down time. The organization spends units only when the compute group is active.
- Organization of data science teams
Depends on whether teams are to share a compute group, whether each team needs its own analytic or analytic GPU compute group
Helps determine the number of analytic compute groups to configure and assign users to
- Number of users who conduct open analytics.
Helps determine the appropriate compute map and cluster scale size.
- Type and volume of analytics workload
Helps determine the appropriate compute map and cluster scale size.
- Deployment plan and analytic workload usage pattern
Helps determine the user accounts and their user profiles, and type of compute profiles to create.
- Users who want to run Deep Learning (DL) or Large Language Model (LLM) workloads need analytic GPU compute groups.
Teradata recommends using the VantageCloud Lake Console to create analytic and analytic GPU compute groups and their profiles compute profiles under these groups. See Admin: Set Up VantageCloud Lake for Bring Your Own LLM/DL Using the Console
SQL: VantageCloud Lake also provides SQL commands CREATE COMPUTE GROUP and CREATE COMPUTE PROFILE Syntax Elements to create analytic or analytic GPU compute groups and profiles for them. These operations must be performed by a system administrator, unless suitable permissions have been granted to a non-admin console user.
To create an analytic compute profile under an analytic compute group, the profile CREATE COMPUTE GROUP Syntax must be specified as 'ANALYTIC'.
To create an analytic GPU compute profile under an analytic GPU compute group, the profile CREATE COMPUTE GROUP Syntax must be specified as 'ANALYTIC_L1'.