Hi!
I had PBcR successfully with demo data. However, when running with bigger data, it is very slow and it seemed only run in one node, so I am doubt whether I am right in configuration. Thus, I look for help here, any suggestion would be grateful!
The command was "PBcR -length 500 -l pbcr -s pacbio.SGE.spec -fastq data.fastq genomeSize=48000000, and when "qstat", I can find the submit job, so it shouldn't run in local model, am I right?
The pacbio.SGE.spec was copied from demo, it has paramters begin with sge, such as sgeScript, sgeFragmentCorrection, sgeOverlapCorrection. However, in the website, they were gridOptionsScript, gridOptionsFragmentCorrection, gridOptionsOverlapCorrection now, did they the same?
Most important, I want to know how to configure these parameters. There are three node with my SGE, two have 64 cpus and one has 32 cpus. Besides, all have 512G memory.
a) I therefor set ovlMemory to 460 that less than 512, should I set it to 460*3? What about ovlStoreMemory?
b) It is said "For other jobs which run as a single multi-core job (Correction, Script) you can set them to utilize a high-memory machine available on your cluster". I think gridOptionsCorrection, gridOptionsScript, gridOptionsFragmentCorrection, gridOptionsOverlapCorrection could use only one node, so I set "orte 32" that consistent with the node with minimal cpus, and "-l mf=460g", am I right?
c) It is said "Make sure that whatever you request in gridOptions for Mhap/Overlap matches what you set for ovlMemory and ovlThreads". I keep the original value of grid (16), so I set ovlThreads to 16, and ovlMemory less than maximal memory (460G). Besides, since "-pe orte" was 16, cnsConcurrency was calculated with (64+64+32)/16, and "-l mf" was calculated with (460*3)/cnsConcurrency=138. I am a idiot here?
d) It is said "tc" could limit concurrently job. I had added tc=10 in pacbio.SGE.spec or "PBcR -tc 20", but all output error, what should I do?
Best wishes!
cat pacbio.SGE.spec merSize=14 # grid info useGrid = 1 scriptOnGrid = 1 frgCorrOnGrid = 1 ovlCorrOnGrid = 1 ovlMemory = 460 ovlThreads = 16 ovlStoreMemory = 460 threads = 16 cnsConcurrency = 10 gridOptionsCorrection=-pe orte 32 -l mf=460g gridOptionsOverlap=-pe orte 16 -l mf=138g gridOptionsMhap=-pe orte 16 -l mf=138g gridOptionsConsensus=-pe orte 16 gridOptionsScript=-pe orte 32 -l mf=460g gridOptionsFragmentCorrection=-pe orte 32 -l mf=460g gridOptionsOverlapCorrection=-pe orte 32 -l mf=460g