In this video from LUG 2015 in Denver, Patrick Farrell from Cray presents: Shared File Performance in Lustre – Challenges and Solutions.
Large scale HPC IO is usually done either with a file per process or to a single shared file. Single shared file IO does not scale well in Lustre compared to file per process. This presentation will give details, examine the reasons for this, and explore existing and potential solutions. Group locks and a new feature, lock ahead, will be discussed in the context of strided IO.”