Phoenix table design

[#1]

I am using Ambari version 2.4.1 with 8 region servers (16 VCPU). Is it ok to have 8*3= 24 buckets for my table? How to upgrade the phoenix to 4.6 where I can create a table and have row timestamp to a Phoenix column? if I create a view on top a table in phoenix, is it going to claim a physical space?

[#2 · Accepted solution]

Please read the follo2ing reference:

http://phoenix.apache.org/performance.html#Salting

For optimal performance, number of salt buckets should match number of region servers). You càn upgrade t HdP 2.5.3 with Phoenix 4.7. Any design decision should be followed by testing.

[#3]

If I create a view on top a table in phoenix, is it going to claim a physical space?

[#4]

No it doesn't (except for metadata). Phoenix doesn't materialize the tables (i.e. physically store the data), but only runs the query mapped by the views.

[#5]

Is it possible to ad index on the views?

Also, I want to add a table with defined column families and few qualifiers. My requirement is to store the data with qualifiers at runtime. How can I make sure to fetch the valid qualifiers while fetching particular record?

[#6]

Yes, index on views work but with some limitations: https://phoenix.apache.org/views.html

I don't have an answer for the second part of your question on cf and qualifiers.

[#7]

for the second question:-

bq. Also, I want to add a table with defined column families and few qualifiers. My requirement is to store the data with qualifiers at runtime. How can I make sure to fetch the valid qualifiers while fetching particular record?

You can use dynamic columns while upserting data and doing SELECT.

https://phoenix.apache.org/dynamic_columns.html