On 08/11/13 03:40 AM, Goor Sasson wrote: > Hi Sebastien & Adrian, > > The Blacklight we consider to use is the one at Urbana Illinois university. > (http://help.igb.illinois.edu/Biocluster#Cluster_Specifications). > A rough estimation for the the total genome size is at least 5000 genomes. > Hopes that these details can shed some light.
Do you expect some sort of power law distribution ? To summarize, you have one sample with 2.5 Gigabases of Illumina data (probably one lane or something like that). And that contains roughly 5000 bacterial genomes. Is that correct ? Also, the next version (2.3.1) will include a patch that will lower the memory usage. see https://github.com/sebhtml/ray/issues/210 > > Thanks, > Goor > > > 2013/11/6 Sébastien Boisvert <sebastien.boisver...@ulaval.ca > <mailto:sebastien.boisver...@ulaval.ca>> > > On 06/11/13 09:49 AM, Goor Sasson wrote: > > Dear Mailing-List, > > > Hi, > > > > I've chosen to use Ray Meta in order to assemble reads from a > pooled metagenome that contains 2.5g of illumina (HiSeq2500) paired-end reads > (read length 100bp, totalling ~250g bp). > I need an gross estimation for the predicted execution time. The > computational platform is a Blacklight platform with ~400 cores (Xeon X7542) > with 2TB RAM. > > > Is it this Blacklight: > http://www.psc.edu/index.php/__computing-resources/blacklight > <http://www.psc.edu/index.php/computing-resources/blacklight> ? > > What is the interconnect (the network) ? > > > Thanks > > > As Adrian Pelin pointed out, it depends a lot of the genomic content of > your metagenome. > > In our paper http://genomebiology.com/2012/__13/12/R122 > <http://genomebiology.com/2012/13/12/R122> , we used 128 machines each with 8 > cores > (Xeon X5560). Each machine had 24 GiB of RAM memory, so that was around 3 > TiB of distributed memory. > It took ~15 hours to assemble metagenome containing 1000 bacterial > genomes (power law abundances) > + human contaminations. Note that this 15 hours included also the > taxonomic profiling and so on. > Doing only an assembly will be shorter in time. > > Obviously, a test that you can you before going for the whole thing is: > > > mpiexec -n 400 Ray -test-network-only -o NetworkTest-1 > > > Then, check the measured latency in NetworkTest-1/NetworkTest.txt > > > > ---Sébastien > > ------------------------------------------------------------------------------ November Webinars for C, C++, Fortran Developers Accelerate application performance with scalable programming models. Explore techniques for threading, error checking, porting, and tuning. Get the most from the latest Intel processors and coprocessors. See abstracts and register http://pubads.g.doubleclick.net/gampad/clk?id=60136231&iu=/4140/ostg.clktrk _______________________________________________ Denovoassembler-users mailing list Denovoassembler-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/denovoassembler-users
