Parallel I/O allows each processor in a multi-processor system to read and write data from multiple processes to a common file independently.

Data-intensive scientific applications use parallel I/O software to access files. In HPC, increasing demands in the I/O system can cause bottlenecks. Parallel I/O plays a fundamental role to balance the fast increase in computational power and the progress of processor architectures.

• Establish the elements of a parallel I/O system:
  • Collective operations.
  • User-defined datatypes to describe both memory and file layout.
  • Communicators to separate application-level message passing from I/O-related message passing.
  • Non-blocking operations.
• Use proper parallel I/O to overcome I/O bottlenecks in HPC environments.

• Use parallel I/O libraries.
• Assess the implications of parallel I/O on application efficiency.
• Implement an application that utilizes parallel I/O to store, retrieve and analyze data.
• Improve the I/O throughput with a parallel I/O file system.
• List I/O system simulators developed for parallel I/O.

• K1.3.2.5.1-B MPI-IO
• K1.3.2.5.2-B PNetCDF