# K1.3.2.5-B Parallel I/O
# Background

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.

# Aim

  * 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.

# Outcomes

  * 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.

# Subskills
  * [[skill-tree:k:1:3:2:5:1:b]]
  * [[skill-tree:k:1:3:2:5:2:b]]