These card sorting examples illustrate several parallel concepts:
  • Increasing the number of parallel workers can decrease the amount of time needed to reach a solution or allow you to tackle larger problems.
  • There is a limit on the number of parallel workers that can share a resource. The contention for a resource increases as more workers try to use it.
  • Independent workers that share no resources still need some amount of coordination to produce coherent results (communication overhead).
  • After a point, additional workers do not speed up a workflow. In fact, the communication overhead and resource contention can make the overall workflow slower.
  • Parallel workflows need to account for the extra work required to coordinate parallel workers.
  • Thoughtful design reduces communication overhead and is more amenable to scaling up to larger problems.
 
©   Cornell University  |  Center for Advanced Computing  |  Copyright Statement  |  Inclusivity Statement