...
Each Worker handles a single compaction task. A compaction is a MapReduce job with name in the following form: <hostname>-compactor-<db>.<table>.<partition>. Each worker submits the job to the cluster (via hive.compactor.job.queue if defined) and waits for the job to finish. hive.compactor.worker.threads determines the number of Workers in each Metastore. The total number of Workrs Workers in the Hive Warehouse determines the maximum number of concurrent compactions.
...