Welcome to SC Cluster by Stanford Computer Science¶
The SC compute cluster, originally the SAIL (Stanford AI Lab) compute cluster, aggregates research compute servers from various research groups within Stanford Computer Science and organized in a grid system. An open-sourced workload/resource scheduler SLURM is exclusively responsible for managing resources such as CPU, Memory and GPU for each job.
The cluster allows both interactive-job and batch-job to be submitted. Interactive job will drop user into a shell and it's a great way for prototyping larger jobs and/or debugging them, whereas batch job can be used to queue up multiple workloads and have them run as soon as compute resources are available, it is the most effective way to use the cluster.
SC cluster is currently on its third iteration, the core servers (sc/scdt) as well as all compute nodes have been upgraded to Ubuntu 24.04 LTS. Please always compile your application and virtual-environment on your respective group's compute node, NOT sc/scdt) to ensure your workload functions correctly.
Eligibility¶
SC Cluster is available to all Stanford Computer Science affiliate who have their server machines managed by CS-IT. Please contact us at action@cs.stanford.edu if you are interested in joining the cluster.
CSID required
Please apply for a CSID Here before requesting access to the cluster
Access Request¶
To request access to the cluster, please fill out the Access Request Form. You will need to fill out the correct group affiliation and contact for your sponor (usually your advisor), please add a note if you are being co-advised by multiple faculty and/or need multiple group affiliation. (for access to multiple partitions and storage systems)
Update Group Affiliation¶
Please fill out the Group Affiliation Update Form when you need to change/add group affiliation, most commonly when you are rotating through different groups.
Help¶
CS-IT is responsible for managing the cluster, please contact us at action@cs.stanford.edu. It's important to be very detail in describing your problem and include SLURM Job ID and other details so we can get back to you quickly and pr