I have been doing an assembly using four lanes of Illumina Hiseq. I have 300bps, 800bps, 8kb and 20Kb insert libraries. Genome size is 80 MB and genome coverage of reads is more than 350x.
By calculating the RAM requirement for this assembly, it turned out that I need around 700GB of RAM for single assembly. I have been submitting velvet jobs on a cluster using PBS scripts, demanding for 32 nodes and 12 processors. Since, 24GB of RAM has been attached per node. If I calculate 32*24 it is equal to 768 GBs, which looks fine for the job submission. But, unfortunately all of my jobs are getting struck after generating velveth results.
My question is, in this kind of run will velvet use whole 768 GBs RAM on total nodes or will it use only 128 GBs ob a single node? I would really appreciate your response.
Thank you so much in advance!
Best regards, Rahul