next up previous contents
Next: Nested model performance Up: The NEMO dCSE Project Previous: The NEMO dCSE Project   Contents

I/O performance

The way in which data is currently input/output to NEMO is not considered ideal for large numbers of processors. As researchers move to use increasingly more complex models at high spatial resolutions larger numbers of processors will be required and this potential I/O bottleneck will therefore need to be addressed. WP1 involves investigating the current scaling and I/O performance of NEMO and identifying methods to improve this via the use of lossless compression algorithms within the NEMO model.