Re: [Denovoassembler-devel] Quick update
Ray -- Parallel genome assemblies for parallel DNA sequencing
Brought to you by:
sebhtml
From: Sébastien B. <seb...@ul...> - 2014-05-16 14:13:43
|
________________________________ > From: cp...@cr... > To: seb...@ul... > CC: jac...@cr... > Date: Fri, 16 May 2014 08:05:55 -0400 > Subject: Quick update > > Hi Sebastien, > > OK, thanks to your help I have Ray running on the XC30 here at Cray. > As I mentioned yesterday I submitted HiSeq2500 with a limit of 12 hr > and using slightly more than 300 cores. This morning ran out of time. > Now, while I start becoming familiar with Titan I’ll continue running > on the XC30 here. Do you have any feeling for how long is going to > take? What I found while experimenting on Titan was that the I/O code that loads sequence from the file system into memory is a bottleneck. Another bottleneck is the code the merge the assembled sequences. > Show we try to run with the restart option? I need to read the > Ray manual to see if you have any info about that option. The option is -read-write-checkpoints Checkpoints It saves the progression of the calculation at key steps. > > In addition, I built the code with a profiler to start familiarizing > myself with the code and try to identify where the bottlenecks are. OK You can also send signal SIGUSR1 to a Ray process to activate RayPlatform debug mode. This option is available by default, but I don't know if you can actually send a signal to a running process on Cray compute nodes. > > Cheers > > carlos |