Assignment 5 (30 points)

In this assignment, each team will extend and redesign the cluster created in Assignment 4 to include a SLURM scheduler.

Part I: (20 points)

You are to augment your cluster from assignment 4 with a SLURM scheduler in the following manner:

• SLURM's slurmctld daemon is to be run on the head node.
• SLURM's slurmd deamons are to be run only on the compute- nodes.

• SLURM's slurmdbd database is to be set up on the metadata node.

• To test your setting, you will demonstrate that you can submit a SLURM batch script that requests 2 compute nodes, compiles hello.c, and run the resulting binaries on the two compute nodes.

• It is recommended that you scale your cluster back down to only 3 compute nodes first

Part II: (10 points)

Once your cluster is working properly, modify your profile as follows:

• Increase the number of compute- nodes from 3 back to 12.
• Modify your batch script such that it requests the necessary resources to perform the timing evaluation of the compiled pi_mc.c source code for the following amount of processes: 2, 4, 6, 8, 10, and 12.

Submission:

You should submit a text file to D2L containing the following:

• URL to your team's Github repository.
• URL to your CloudLab experiment that has a successful deployment of the profile. The experiment's duration should have been extended such that there are at least 24 hours remain from the D2L submission time.
• Information about the directory containing the job output for Part II.
• Any additional notes that I must be aware of when testing your experiment
• This is a team assignment. Each team should only have one submission

Technical Reference:

You can start with https://slurm.schedmd.com/quickstart_admin.html