@mquresh Thank you very much. You are right, I only set mapreduce.job.queuename, but not set the tez.queue.name, so It only use the default queue which only using 30% computational resource using Hive on Tez. ... View more