Why dose Jaffa take 28 h for 47M paired-end reads of 300 length (150-150) in 'DIRECT' mode?

[30ma]

Hi 

I am running jaffa in a nextflow pipeline and  it take very long for each sample. 

I used  18 cpus as #threads passing to bpipe. 

It took 28 hours for 47M reads...

According to the trace memory should not be a problem. However it seems to be very poorly parallelized.

CPU: 153.7%
peakRSS: 8.5 GB
peakVMEM: 40.4 GB

Dose it look as it expected or I can improve it in a way (like increasing the number of threads)? 

cheers,  

Sima

[Nadia Davidson]

Hi Sima,

Yes this looks as expected. Unfortunately JAFFA is slow and not parallelized. The threads passed to bpipe in direct mode don't actually do much. On the up side, this might mean you can run more of your samples in parallel, which is what we tend to do. We are working on a new version which is parallelized (and much faster even for 1 thread) and hope to have that out later this year. 

Cheers,

Nadia.