Scalable I/O of large-scale molecular dynamics simulations: A data-compression algorithm

Disk space, input/output (I/O) speed, and data-transfer bandwidth present a major bottleneck in large-scale molecular dynamics simulations, which require storing positions and velocities of multimillion atoms. A data compression algorithm is designed for scalable I/O of molecular dynamics data. The algorithm uses octree indexing and sorts atoms accordingly on the resulting space-filling curve. By storing differences of successive atomic coordinates and using an adaptive, variable-length encoding to handle exceptional values, the I/O size is reduced by an order-of-magnitude with user-controlled error bound.