Spotfire: Does a Cluster of TIBCO Spotfire Statistics Services (TSSS) support round-robin scheduling of jobs across its worker nodes?

Does a Cluster of TIBCO Spotfire Statistics Services (TSSS) support round-robin scheduling of jobs across its worker nodes?

A TIBCO Spotfire Statistics Services (TSSS) cluster with at least one Manager node and multiple Worker nodes will route jobs to available worker nodes for execution. From the TSSS jobs URL (http://<servername>:<port>/<ServiceName>/api/v8/jobs), you may see that all or most of the jobs get routed to only a single worker node, without routing jobs to other worker nodes. This may create confusion about whether the TSSS cluster is working correctly, and whether there is a configuration issue that prevents routing jobs to other worker nodes.

Note:  In the above jobs URL, replace <servername>, <port> and <ServiceName> as defined in your TSSS cluster installation.

Resolution

The clustering algorithm used in TSSS does not guarantee round-robin scheduling, but rather will fill one Worker with jobs before scheduling jobs to the next Worker. Clustering provides improved performance.  So the behavior mentioned above is expected, as long as the total number of engines available across the cluster is more than 0 and jobs are not queued extensively.
 
For example, with 2 Worker nodes in a cluster that has 4 engines available on each Worker node, if 4 jobs come in simultaneously, it is not a problem if all 4 of those jobs get allocated to the same Worker node. In this scenario, if 5 jobs come in simultaneously and 4 jobs get allocated to one Worker node, then the remaining (5th) job will get routed to the second Worker node. However, in this case if the 5th job gets queued without being routed to the other Worker node, then that is a cluster configuration issue that needs to be investigated.