ODU Research Computing Forum

Should I connect to Turing or Wahab?

The are two HPC clusters, which one should I use: Turing or Wahab?
If I am on Turing how do I access my data through Wahab?
These are questions often posed to me by new people working in Carpenter Lab.

ANSWER:

I recommend using Wahab for your research. It is a newer and faster cluster. It can perform all of our genomic data handling better than Turing. We use Turing for instructional purposes so that students and workshop participants can practice genomic pipelines on smaller data sets.

Long term information storage, that is information stored on the RC drive, is independent of the cluster you log in to. Your data can be accessed from both clusters. Because we are used to our personal computers, it is difficult for us non-computer scientists to visualize how the cluster works. Imagine that your hard drive is in one corner of a room, your mother board in another and you carried a monitor and keyboard with you. In a sense that is how the clusters work, each piece communicating with the other. Turing and Wahab have a scratch drive that can only be accessed through that cluster. I recommend using your scratch directory as a temporary staging area for your data set as you perform your analysis.

I highly recommend reviewing the documentation for the HPC which will give you have a better understanding of the answers to these two questions.

Very respectfully,

Iván

Welcome Ivan to our forum! The Research Computing Services does recommend newer users to start using Wahab right away, as it is a more recent cluster, with more recent hardware, OS, software and library. Each Wahab compute node contains 384 GB RAM (3x that of Turing) and 40 CPU cores (about 25% more cores per node than Turing).

Many bioinformatics tools require a large amounts of memory. With Wahab, you can fit more data onto the computer, resulting in more efficient processing. Please refer to the documentation of your specific software to know about its memory usage. When using large amounts of memory, please be cognizant that other jobs may be running on the same node, thus causing your job or others’ jobs to fail. The sbatch command provides the --mem switch to let SLURM know that you need that much memory per compute node. Please use that switch so that your jobs are placed on the right resources.

These being said, Turing has a few nodes, located in the himem SLURM partition, that has even more memory (768 GB or 512 GB) compared to Wahab. For jobs that require extreme amounts of memory (e.g. Trinity), the use of Turing is indispensable.