Sigma

Introduction

This page tries to describe the job scheduling policy on Sigma, i.e how the system decides which jobs should be started and when.

The scheduling policy for Sigma is identical to the one on Tetralith (read about it here) with the following exceptions:

The maximum timelimit for a job on Sigma is 3 days (72h)¹
The "devel" nodes on Sigma have a dual purpose. Just as on Tetralith, you may use them for test and development. However, on Sigma you may also use the devel nodes for interactive work where it's important to get access to nodes quickly. Between 08:00 and 22:00 local time (CET/CEST), 8 nodes are available in the "devel" reservation. No user may use more than two of these nodes (or 64 cores) simultaneously.²
The Sigma GPU nodes are in a separate reservation and are only available to GPU projects (project name LiU-gpu-*).
If you need to increase the size of your LiU project, email NSC Support. There are no fixed limits to the size of LiU projects.

This has been decided in dialogue with the LiU Sigma projects, to ensure that queue times does not become too long. Since Sigma only has 108 compute nodes, allowing 7 days jobs would mean that only a small number of nodes become available for new jobs each day, and queue times could sometimes be very high even for high-priority projects.↩
This has been decided in dialogue with the LiU Sigma projects. The idea is to have a relatively large number of devel nodes that can be allocated for a full workday, to better allow for interactive access to resources for users who need it. The number of devel nodes may change over time depending on usage.↩