What's Granted in Plexus
Plexus manages compute resources for customers using the concept of queues, which are based on Slurm partitions.
What is a Queue in Plexus?
A Plexus queue is a logical grouping of compute nodes, derived from the Slurm partition concept. Each queue (partition) has:
- A specific set of compute nodes with defined hardware characteristics (e.g., GPU/CPU type, memory)
- Resource limits and scheduling policies (e.g., workload limits, time limits)
- Intended use cases (e.g., AI/ML, HPC, data analytics)
- Access controls for different user groups or teams
When users submit workloads to a queue, Slurm schedules those workloads on the nodes assigned to that queue, applying all configured constraints. This organization helps manage cluster resources efficiently for various workloads and user groups.
How Queues are Granted
Customers are assigned to a specific team, each with access to a list of associated queues. When creating a workload, customers select from their available queues, and Plexus automatically prepares the environment and resources needed to run the workload in the selected queue.
Tips
- Choose the queue that best matches your workload’s resource needs.
- Review queue limits before submitting large or long-running workloads.