Workflow nodes¶
We know that traditional supercomputing centers (including NERSC) are sometimes not ideal for running long-lived scientific workflow tools. Batch jobs and queue wait times can be obstacles to effectively using these tools.
We understand these difficulties! In our effort to better support workflows, Cori has two dedicated nodes especially for running workflow software: cori20 and cori21.
Workflow nodes are for orchestrating workflows
NERSC workflow nodes are for coordinating your workflow, not running your workflow. What does this mean? Having a cronjob that wakes up once an hour, checks for data, and launches jobs on the compute nodes is an excellent use case for the workflow nodes. Actually running your computations on the workflow nodes is prohibited.
Like other login nodes at NERSC, the workflow nodes are shared resources. Please adhere to the NERSC login node guidelines and be a good citizen.
How can I get access to the Cori workflow nodes?¶
Please submit a ticket at help.nersc.gov
To help us decide if your use case is appropriate for the workflow nodes, you will be asked:
User:
Email:
Repository Name:
Purpose:
Estimated memory usage:
Estimated CPU usage:
Estimated Data usage:
Estimated I/O:
Frequency / length of process:
Need external resources:
Workflow node details¶
cori20 and cori21 are similar to the rest of the NERSC login nodes. They are Haswell nodes with 500 GB of memory per node. Once you have logged onto Cori, you can reach them via
ssh coriwork
A load balancer will then direct you to either cori20 or cori21. From Cori, you can also ssh directly to cori20 or cori21.