Cori s lustre scratch file system provides more than 700 gbs of peak bandwidth and 30 pb of disk capacity, compared to about 150 gbs and 7 pb on edison, a cray xc30 based on intel xeon ivy bridge processors. Cori has a dedicated, large, local, parallel scratch file system based on lustre. Nersc has also added software defined networking features to cori to more. Cori scratch cori scratch is storage space for each user located on a lustre filesystem accessible from cori and cori exvivo. Large io operations should always be directed to scratch filesystems. Cori has dedicated large, local, parallel scratch file systems. Cori offers additional features and capabilities that can be of use to jgi researchers. A scalable tool for deploying linux containers in highperformance computing. The scratch file systems are intended for temporary uses such as storage of checkpoints or application input and output. Cori scratch cori scratch is a lustre filesystem designed for high performance temporary storage of large files. The data transfer nodes are nersc servers dedicated to performing. It is intended to support large io for jobs that are being actively computed on the cori system.
We recommend that you run your jobs, especially data intensive ones, from the cori scratch file system. Pdf file system monitoring as a window into user io. The first time you try to connect from a nersc system cori, dtns, etc. Cori scratch is a lustre filesystem designed for high performance temporary storage of large files. Examples of intended usage would be running python scripts to download data from a. Scriptsbenchmarks for running tensorflow distributed on cori. Data and io intensive applications should use the local scratch or burst buffer filesystems. File system monitoring as a window into user io requirements.
If you want to tweak things you can build your own conda environment you can start by cloning ours and update the setup. Cori supercomputer now fully installed at berkeley lab. Cori is a cray xc40 with a peak performance of about 30 petaflops. The examples here utilize a custom conda environment on cori scratch which should be fine for you to use for now.
Note that slurm currently only supports jobs which use the same number of processor per process. These buttons allow you to manage notebook servers running on cori or in spin. Introduction to nersc resources berkeley lab computing sciences. National energy research scientific computing center. One fdr ib connection to cori scratch globalcscratch1. When you log into jupyterhub at nersc, you will see a console or home page with some buttons. Use the the unix command cp, tar or rsync to copy files within the same computational system. The system is named in honor of american biochemist gerty cori, the first american woman to win a nobel prize and the first woman to be awarded the prize in physiology or medicine. The scratch file system is intended for temporary uses such as.
A scalable tool for deploying linux containers in high. It is intended to support large io for jobs that are being actively. Cori is comprised of 2,388 intel xeon haswell processor nodes, 9,688 intel xeon phi knights landing knl nodes. Documentation and examples for using slurm at nersc can be found here. In addition to a broad survey of the nersc hpc population, we follow the evolution of. Threads per process how many threads per process do you want to use. Spawns jupyter notebooks on specialpurpose largememory nodes of cori cori, cori14, cori19 exposes gpfs and. Usually thats the local scratch file system or the burst buffer.
560 485 417 205 1371 714 1589 443 1002 1475 590 1272 781 673 1231 845 1309 1475 1040 64 701 937 1112 61 216 1268 354 1110 637 1091 1131 426 710 296 892 423 90 1352 989 108 1403 1081 279 1363