• Improving performance sometimes involves trade offs between time and space:
  • faster access time by using more storage space, for example storing sparse structures by ignoring their sparseness
  • less storage space with slower access time, for example packing ragged array in single list with separate index marking where each row starts
• If disk or network bandwidth is the bottleneck, smaller can also be faster.
• See tomorrow's sessions on netCDF-4 for time and space performance improvements made possible by the use of HDF-5 as a storage layer.

2010 Unidata NetCDF Workshop > Formats and Performance