What's a Plexus Workload
Plexus manages compute resources through the concept of workloads, which correspond to Slurm jobs executed on the underlying cluster.
What is a Slurm Job?
A Slurm job is a unit of work submitted by a user to a Slurm-managed cluster for execution. It typically consists of a script or command that specifies the tasks to be performed, along with resource requirements such as the number of CPUs, GPUs, memory, and runtime limits. When a job is submitted, Slurm schedules it to run on available compute nodes that meet the requested criteria. Jobs can be run in batch mode (non-interactive) or interactive mode, and their progress and status can be monitored through Slurm commands or the platform interface.
When submitting jobs to a Slurm-managed cluster, users must have access to the cluster through an SSH connection.
What is a Plexus Workload?
A Plexus workload is a unit of work submitted by a user to either a Slurm-managed or Kubernetes cluster. For each application, Plexus offers a default resource configuration, which users can customize to fit their specific requirements. Workloads can be executed in either batch (non-interactive) or interactive mode.
Plexus handles the allocation of resources needed to execute the application in the cluster. It monitors the progress and status of each workload and provides tools for users to interact with their running workloads.
Ultimately, a workload is the execution of an application within an isolated environment on a cluster.
With Plexus, users do not need special access to the underlying clusters. One Plexus account provides access to all types of compute resources.