Hilbert curve scheduling

Hilbert curves.

In parallel processing, the Hilbert curve scheduling method turns a multidimensional task allocation problem into a one-dimensional space filling problem using Hilbert curves, assigning related tasks to locations with higher levels of proximity.[1] Other space filling curves may also be used in various computing applications for similar purposes.[2]

The SLURM job scheduler which is used on a number of supercomputers uses a best fit algorithm based on Hilbert curve scheduling in order to optimize locality of task assignments.[2]

  1. ^ Scheduling for Parallel Processing by Maciej Drozdowski 2009 ISBN 1-84882-309-6 page 166
  2. ^ a b Job Scheduling Strategies for Parallel Processing: by Eitan Frachtenberg and Uwe Schwiegelshohn 2010 ISBN 3-642-04632-0 pages 138-144