Running Blast on RCC (Updated)

[tabinks]

See #29.

[tabinks]

Log in to RCC

• From a terminal login using ssh

ssh <cnetid>@midway.rcc.uchicago.edu

Download Files

• Query sequence file

wget http://uchicago.bio/sessions/session4/protein1.fasta

• Sbatch script to launch a blastp job on a single node. Note that it uses a database that is located on a shared directory on RCC.

wget https://raw.githubusercontent.com/uchicago-bio/RCC-Utilities/master/single-node-blast.sbatch

Run the Script

From the command line:

sbatch single-node-blast.sbatch <some-unique-text-to-identify-this-job>

Adjust the number of CPUs per Task

This task is running on a single node, but we will vary the number of cpus that are assigned to each task. Do runs with 1,2,4,8,16 cpus per task. The Sandy Bridge chips (sandyb) are the default chips our allocation has. It has 16cores and 32G of memory. If you need more memory (you may get a warning from SLURM), you should investigate the the bigmem partition and add more memory per cpu.

#SBATCH --nodes=1    # This is the number of nodes we are requesting for the job
#SBATCH --tasks=       # The number of tasks this job will have
#SBATCH --ntasks-per-node=1  # The number of tasks that will run on each node
#SBATCH --cpus-per-task=16    # Number of cpus per task
#SBATCH --partition=sandyb     # The chip type you are requesting
#SBATCH --mem-per-cpu=2000  # Some jobs will require more memory 

[tabinks]

If you downloaded the Jupiter Notebook, please note that I have added the instructions to match above, but the questions are the same.