Cost Saving Plan

Cost-saving plans estimate the savings you can achieve by utilizing pre-calculated versions of your metrics. You can generate a cost-saving plan for a query audit group with common computation logic: if there is no common computation for a group, you cannot generate a cost-saving plan for that group. Perform this action by choosing the "Generate Cost Saving Plan" action on an audit group's detail page.

📘

Note

Note that cost statistics can only be displayed when the query collection being audited is sourced from a query history table in your warehouse. Cost plans cannot be displayed when the query collection being audited is sourced from a CSV or compliant table.

Run Options

We create a temporary table based on the common computation logic when you generate a cost-saving plan. If you choose the "dry run" option, then:

  • we perform a dry run of the original queries in the group that match the common logic
  • we perform a dry run of the same queries rewritten to use the common logic (temporary table)
  • we then compare the performance of the rewritten queries to the original queries

If you choose the "real run" option, then:

  • we run the original queries in the group that match the common logic
  • we run the same queries rewritten to use the common logic (temporary table)
  • we then compare the performance of the rewritten queries to the original queries

📘

Dry run or real run?

A dry run is cheaper in BigQuery since you are not charged for performing a dry run; however, a dry run can only estimates the number of bytes read by the query. Other statistics like query duration are not available

Results

The way to read a row in the "Rewritten queries" table is the following:

  • For the query with index 123 in the query collection that was audited:
    • the Duration of the original query was 248 ms
    • the Duration of the rewritten query was 222 ms
    • the rewritten query was 222 - 248 = -26 ms faster, so the duration of the rewritten query is 11% faster than the duration of the original query
  • Processed Bytes and Output Rows are interpreted in the same way as Duration:
    • the original query output 4 rows
    • the rewritten query output 4 rows
    • the rewritten query output 4 - 4 = 0 additional rows, for a change of 0%

The "Savings Summary" stats take the max (average) across the values in each row. For example, if one row has (-10%) and one row has (-6%), then the average across the two rows is -8%. Note that changes between the original and rewritten queries can be noisy for small queries (short duration or few processed bytes).

BigQuery FAQ

  • Temporary tables that materialize the common computation logic are stored in a singleorigin schema that we create in the project you have connected to.
  • When you perform a "dry run" report, we will only surface the Processed Bytes stat provided by BigQuery - not all stats are available since queries are not executed.
  • Why don't I see the option to generate a cost-saving plan for a group?
    • If there is no common computation for a group, you cannot generate a cost-saving plan for that group.

Snowflake FAQ

  • What permissions does my Single Origin connector need to automatically generate Cost Savings Plans?
    • For complete automation, the role you are using for your Single Origin connector will need ACCOUNTADMIN level permissions.
    • If you do not want to give your Single Origin connector will need ACCOUNTADMIN level permissions, then you will need to create a Database called SO_DB and Schema called SINGLEORIGIN in your Snowflake environment. This is where temporary tables that materialize the common computation logic are created during report creation.
  • Temporary tables that materialize the common computation logic are stored in a schema called SO_DB.SINGLEORIGIN in the Snowflake environment you connect to.
  • Why don't I see the option to generate a cost-saving plan for a group?
    • If there is no common computation for a group, you cannot generate a cost-saving plan for that goup.

Updated 6 days ago