How clustering affects priority and worker selection for jobs

A diagram showing how the same 4 jobs have their priority sorted differently depending on which cluster the job and worker are each in.

Workers in the / cluster (also known as the 'root')

Jobs are sorted by

1. priority
2. then by job id (first come, first served - FCFS)

Workers in cluster /A

1. jobs in cluster /A are considered first, sorted by
   
   1. priority
   2. then by job id - FCFS
2. cluster /B jobs are considered next, sorted by:
   
   1. priority
   2. then by job id - FCFS

Workers in cluster /B

1. jobs in cluster /B are considered first, sorted by
   
   1. priority
   2. then by job id - FCFS
2. cluster /A jobs are considered next, sorted by:
   
   1. priority
   2. then by job id - FCFS

Custom Queuing Algorithms

Although Qube is shipped with several pre-built and optimized queuing algorithms, Qube also allows developers to take advantage of the research PipelineFX has done in queuing system design to write their own queuing system algorithm. This feature enables a site to design an queuing algorithm tailored to its own policies.

See the Custom Queuing Algorithms in the Developer's Guide for more information on how to develop custom queuing algorithms.