What is Panda?

Panda is a parallel I/O library developed at Professor Marianne Winslett's group in the University of Illinois, Urbana-Champaign. It is a easy-to-use, portable, high-performance I/O library which targets distributed and distributed-shared memory multiprocessors such as supercomputers and clusters of workstations. Panda supports parallel I/O for large-scale multidimensional arrays, which are the most frequently used data structure in scientific and engineering applications. It provides high-level array I/O interface for single-program multiple-data (SPMD) style programs. I/O operations supported by Panda are collective, where all the compute processors cooperatively perform I/O. Array snapshot and checkpoint/restart operations are supported in Panda. More information on Panda can be found at http://drl.cs.uiuc.edu/panda.