I'm analyzing my BQ project to see if switching to Physical Bytes Pricing model will be beneficial and I ran the query that was recommended in here and found that despite great compression ratios in our datasets, we still are having to to pay-up when we switch from logical to physical pricing model.
The primary reason I found was that time-travel bytes are way higher in our datasets for certain tables. For physical bytes pricing model, time-travel bytes are charged $. A lot of tables that are being built in our env are CREATE OR REPLACE TABLE SQL syntax, which might be prompting the time-travel feature to save the whole table as backup. What are some optimizing changes I can make to reduce time-travel bytes. Some I could think of are -
1. Make sure of TEMP tables when the table is just an intermediary result table that are not used outside of the multi-query job.
2. May be delete the table and then CREATE it again ? instead of create or replace table syntax ? Am not sure.
could anyone suggest any optimizations that I can do to reduce time-travel bytes ? in this case or in general.
LinkedIn
Twitter
