Can we get better performance for hive queries by using SSD?

[#1]

One of my client is using Azure based IaaS for their HDP cluster. They are open to using more expensive storage to get better performance.

Is it recommended to use SSD for some of the data in hive tables, to get that boost in performance? Also what are the steps to make your temporary storage to point to SSD, that is used by Tez/MR jobs?

[#2]

I understood that yarn.nodemanager.local-dirs would be the setting to point to SSD to get better performance on shuffle and other temporary usage. I also would like to confirm it.

[#3 · Accepted solution]

This thread has all the details needed:

http://community.hortonworks.com/questions/1405/can-you-please-advise-about-how-best-to-use-this-s.h...