I'm wondering when should we use Tez or MR as an execution engine for our queries running in hive?
Slide 9 and later from the following link can give more details: http://www.slideshare.net/Hadoop_Summit/w-235phall1pandey As one of the advantage of Tez is that it avoids multiple writes to HDFS which definitely improves overall performance.
View solution in original post
I would say as the latest release of HDP, I see very little to any reason to use MR over Tez. I would say default to tez and use MR if and when required (not may use cases).
This is a fine answer that lists other aspects of considering Tez over MR for Hivehttp://community.hortonworks.com/answers/83488/view.html