The International Conference for High Performance Computing, Networking, Storage and Analysis
Graphical Analysis of I/O in High Performance Computing.
Student: Denver Coker (Oak Ridge National Laboratory)
Supervisor: Jeremy S. Logan (Oak Ridge National Laboratory)
Abstract: In high performance computing, full simulation application jobs can be time consuming and expensive to run. To analyze I/O performance a skeletal application is used to monitor only the I/O, thereby making the simulation less costly in many ways. After conversion, the output is then imported into graphical analysis software such as R for analysis. The script I have written takes all possible combinations of the variables from the skeletal application, and prints only the significant graphical combinations. Also the script will automatically color the graph intelligently as well. This method of data analysis is much more efficient since every graphical combination is not outputted but only the ones the user is interested in.