Index
SOCA is supporting OpenPBS since the initial release and has added support for additional schedulers and workload managers since version 25.11.0.
Important
As of today, only OpenPBS has been fully battle-tested and is considered production-ready.
Integration work for other schedulers is ongoing, and their basic functionality has been validated. However, we recommend waiting until their status is marked as Stable before using them for full production workloads.
| OpenPBS | IBM LSF | SchedMD Slurm | AWS Elastic Kubernete Service (EKS) | |
|---|---|---|---|---|
| Status | ✅ (Stable and tested) | ⚠️ (Preview mode) | ⚠️ (Preview mode) | ⚠️ (Preview mode) |
| Automated Installation (Server) | ✅ | ✅ | ✅ | Connect to existing EKS |
| Automated Installation (Client) | ✅ | ✅ | ✅ | ✅ |
| On-Demand Nodes Provisioning | ✅ | ✅ | ❌ | ✅ (via EKS) |
| Always On Nodes | ✅ | ⚠️ (in dev) | ❌ | N/A |
| Job Shared | ✅ | ⚠️ (in dev) | ❌ | N/A |
| Support for SOCA Job Resources | ✅ | ✅ | ❌ | N/A |
| List jobs via web interface | ✅ | ✅ | ✅ | ✅ |
| Delete jobs via web interface | ✅ | ✅ | ✅ | ✅ |
| Submit jobs via web interface | ✅ | ✅ | ✅ | ✅ |
| Connect to on-prem environment | ✅ | ✅ | ✅ | ✅ |
| View More | View More | View More | View More |
Please note the following:
- A single SOCA Controller can host multiple schedulers.
- A user can submit Slurm/LSF and/or OpenPBS jobs on the same machine, as long as all schedulers are registered on SOCA
- You can add or remove schedulers from your SOCA configuration dynamically, without restarting any services.
- SOCA can automatically provision compute nodes based on job submissions using the using the SOCA Job Resources Parameters
- These parameters are supported by all scheduler providers, via
-lfor OpenPBS,-jDfor LSF and--commentfor Slurm - Automatic Host Provisioning and SOCA Job Resources Management is only supported on LSF and OpenPBS
- You can connect to your AWS Parallel Computing Service (PCS) environments
SOCA is capable of managing node provisioning or serving as an HPC client to interface with your on-premise schedulers.
For example, you can set up LSF on SOCA to enable cloud bursting with automatic node provisioning, while simultaneously configuring SOCA to act as a client for your on-premises LSF cluster.
This setup allows end-users to view jobs from all HPC environment environments on a single page through the SOCA My Jobs dashboard.
You can get detailed job information or kill a job using the same web-interface
Example of Job Submission with SOCA resources:
# OpenPBS/PBSPro - via the job resource list
qsub -l instance_type=m6i.xlarge -l scratch_size=500 ...
# LSF - via the job description
bsub -Jd "instance_type=m6i.xlarge scratch_size=500" ...
# Slurm - via the job comment section
sbatch --comment "instance_type=m6i.xlarge scratch_size=500" ...