Researcher Reduces File Transfer Times from 61 hours to 21 minutes
Craig Mattocks is a climate researcher at the University of Miami. Craig writes:
"Globus makes OLAM global climate simulations manageable! I just used Globus/Globus Connect to transfer 28 GB of OLAM output (62 history files) from the Trestles cluster with a single click in a web browser at a speed of 183.3 MBits/sec vs. the usual 1 MBit/sec (using sftp/rsync). 20 minutes instead of 61 hours! Another file transfer went even faster, almost 200 MBits/sec! The fastest we have been able to download the data using sftp/rsync is about 1 MBit/sec, so that's a factor 200 speed-up. It took only 5 minutes to move this data from one side of the country to the other (San Diego to Miami) using Globus. This particular transfer would have taken almost 17 hours by sftp/rsync. Amazing! Thank you for developing this new easy-to-use, high-speed transfer tool and making it available for XSEDE users. Great work y'all!
OLAM is a new global climate model with local mesh refinement on hexagonal or triangular finite volume grids. We are running a series of 10-year "timeslice" climate simulations and are working to create a gateway for the model. Each of these OLAM simulations generates about 8 TB of data. We were really struggling to keep up with the volume of data. This new file transfer capability makes managing this data possible for local archival and post-processing afterward. In the future I imagine we will include Globus in the OLAM gateway somehow, so users (students, teachers, researchers) will be able to download their model results to a local disk."