locked
PDW Statistics on distributed tables (hash candidate columns) RRS feed

  • Question

  • I am doing some performance optimisation on my SQL statements in a PDW. I am joining several tables which are hash distributed. Distribution on the tables in question is based on the joining columns that I’m using. However, it appears like those columns do not have any statistics on them. I wanted to know if it is necessary to explicitly create statistics on those columns. One of my colleagues thinks that since the columns in question are already used on distribution (i.e. they are hash candidate columns), it’s not necessary to add another layer of statistics on them. What’s the best practice in this situation? Personally, I feel that creating statistics on those columns will enhance performance even more. I need your advice.

    Kind regards,


    Mpumelelo

    • Edited by Mpumelelo S Thursday, March 16, 2017 12:07 PM
    Thursday, March 16, 2017 11:10 AM

Answers

All replies