Scheduler Overview
HPC Cluster Service supports Fsched scheduler clusters (Fsched clusters) and Linux clusters without a scheduler (None Linux). Fsched clusters are suitable for complex scenarios requiring high concurrency, high performance, and large-scale computing. None Linux clusters are more suitable for small-scale data processing, development/testing, and scenarios that do not require high concurrency or large scale.
Feature comparison by scheduler type
Dynamic nodes and autoscaling are supported only after Hybrid Cloud is enabled in FCP-Suite.
| Category | Item | Fsched cluster | None Linux |
|---|---|---|---|
| Cluster config | Cluster name | ✓ | ✓ |
| Per-user resource restriction | ✓ | ✗ | |
| SSH login restriction | ✓ | ✗ | |
| Alert service | ✓ | ✓ | |
| Custom parameters | ✓ | ✗ | |
| Release protection | ✓ | ✓ | |
| Mount configuration | ✓ | ✓ | |
| Compute partitions | Multiple partitions | ✓ | ✗ |
| Same compute node in multiple partitions | ✓ | ✗ | |
| Partition name | ✓ | ✓ | |
| Swap configuration | ✓ | ✓ | |
| Tags | ✓ | ✓ | |
| Fsched advanced configuration | ✓ | ✗ | |
| Static nodes | ✓ | ✓ | |
| Dynamic nodes | ✓ | ✓ | |
| Autoscaling | ✓ | ✗ | |
| Login partition | Static nodes | ✓ | ✗ |
| Dynamic nodes | ✓ | ✗ | |
| Head partition | Static nodes | ✓ | ✗ |
| Dynamic nodes | ✓ | ✗ |
Note:
When using Fsched autoscaling, note that srun is an interactive job. During instance creation and initialization, network connectivity may be unstable and can cause submission errors. To avoid this issue, we strongly recommend submitting jobs with sbatch.