Anomaly Detection
Perks
Powered By GitBook
BigQuery FinOps Dashboard
Your Swiss army knife for highlighting inefficiencies in your organization's BigQuery usage

Overview

BigQuery FinOps is your Swiss army knife for highlighting inefficiencies in your organization's BigQuery usage, and its insights are displayed in the BigQuery FinOps Dashboard, located in your Cloud Management Platform. Without configuring anything, you'll have the most important BigQuery insights brought to you.
To access the BigQuery FinOps Dashboard, you must verify that your Google Cloud service account was successfully uploaded and granted the appropriate permissions. Read more on linking your Google Cloud Organization.
If successfully uploaded, you will see:
    Either a Healthy or Partia value underneath the Status column in your Configured Accounts widget, and
    Healthy status next to "BigQuery FinOps" in the Features widget.
    [Optional] Healthy status next to "BigQuery FinOps Advanced" in the Features widget
A screenshot of a BigQuery FinOps dashboard

Setup

Once the Google Cloud Service Account setup is complete, click the **'**Attach' button and choose the BigQuery FinOps dashboard from the list.
A screenshot showing the location of the Attach icon
Cloud Management Platform will now start gathering the information on your usage patterns.
The BigQuery historical jobs are backfilled for the last 30 days
While this information is being collected, your BigQuery FinOps will have the three following states:
1) We have started to analyze your historical usage. So far we have processed X%. In the meantime, you can explore our training Perks!
A screenshot of the notification text above
2) The scan is completed, and X% of your data was already processed. You are almost there!
A screenshot of the notification text above
3) BigQuery FinOps highlights inefficiencies in your BigQuery usage. The analysis represents the last 30 days of use.
Once available, you'll see the dashboard appear with the statistics of your organization's BigQuery usage.

Understanding the BigQuery FinOps Dashboard

The BigQuery FinOps dashboard is compiled of widgets that will provide you more insight into your organization's statistics.
A screenshot of the widgets on a BigQuery FinOps dashboard

Elements of the Dashboard

Below is a list of the elements that make up your BigQuery FinOps dashboard, with an understanding of their purpose:
    1.
    Timeframe - all insights displayed are from the last 30 days.
    2.
    BigQuery Spend by SKU - cloud cost and usage analytics
    3.
    BigQuery Recommendations- comprehensive recommendations with further details on how to act on each recommendation.
    4.
    BigQuery Explorer- get more granular information as far as your team's BigQuery usage, completely modifiable by the user in the following format:
Top 10 (Project, Billing Project, Dataset, Table, User) by (Scan Price, Scan TB, Storage Price, Storage TB).
Clicking on one of the results in the BigQuery Usage Explorer widget will open a pop-up where you can get even more granular with your usage insights.
5. BigQuery Scans by Table Type - see how much data you are scanning from various types of tables, broken down between unpartitioned tables and various types of partitioned tables, along with external sources such as Google Sheets.
Click on a specific table type to see what are the most frequently-scanned tables within that type.
Note on the BigQuery FinOps Advanced

BigQuery Recommendation Types

Below is a list of the recommendation categories you'll see in the BigQuery Recommendations widget, and how to use the information contained within each recommendation.
Backup and Remove Unused Tables - Backup and remove the unused tables listed under the "Table" column. If the table has multiple partitions, click on the number listed under "Partition(s) to Remove" to see precisely which partitions should be removed.
Cluster your tables - Cluster the tables listed under the "Table" column by the field(s) suggested under the "Cluster By" column.
Enforce Partition Fields - Use the suggested partitioned fields(s) under the "Partition Field" column for the corresponding queries listed under the "Query ID" column.
Partition your tables - Partition the tables listed under the "Table" column by the suggested field(s) listed under the "Partition Fields" column.
Limit query jobs - Reduce job execution frequency of the listed jobs under the "Query ID" column by the percentage you choose on the slider, and view the associated savings of each reduced job under the "Savings by Reducing Jobs" column.
The BigQuery Finops Dashboard does not currently reflect reservations and your job costs are displayed as on-demand. If you use reservations, please disregard the BigQuery Finops Dashboard's recommendations until we include reservations in our future release.

BigQuery FinOps Frequently Asked Questions (FAQ)

Oftentimes we get asked what the non-read-only permissions are for, and so we'd like to share more about how it plays into the process of creating your BigQuery FinOps Dashboard.

What are the permissions we are referring to?

bigquery.datasets.create, logging.sinks.create, bigquery.jobs.create, and bigquery.tables.getData

Why do you separate between BigQuery FinOps and BigQuery FinOps Advanced permissions?

We require bigquery.tables.getData in order to provide clustering recommendations, and this is separated from the permissions required under the BigQuery FinOps feature.
This allows us to query your BigQuery tables and determine your top 20 non-clustered tables, which field(s) are the best candidates to cluster those tables by, and in what order.
More specifically, this helps us identify the cardinality of the columns in your top 20 non-clustered tables which get referenced in the WHERE clause and are possible to cluster on. Knowing how many distinct elements exist in each column enables us to compute the average chunk size and see what the best candidates in terms of savings would be.**
However, you can still use and benefit from the BigQuery FinOps Dashboard without receiving clustering recommendations. That's why we separate the two.
Given that you execute similar queries as you did during the previous 30 days, and reference fields in the order we recommend.

What datasets are you creating, where and for what purpose?

A dataset called doitintl-cmp-bq is created in the billing project attached to the service account you add to the CMP.

What log sinks are you creating, where and for what purpose?

A sink for query jobs is created in the same project as (1). This sink pushes all your bigquery jobs into a table under <PROJECTID>.doitintl-cmp-bq. cl oudaudit_googleapis_com_data_ access

What queries will you be running, where, and for what purpose?

We have two main processes, once of which runs queries and the other not: 1) Enrichment process: We take the data from doitintl-cmp-bq. cloudaudit_ googleapis_com_data_access and create a new table doitintl-cmp-bq.enrichedJobs. This is done at no cost to you, but rather an algorithmic and API based enrichment from our side to have a clean dataset to work with in the processes that follow. 2) BigQuery Finops process: We create a few UDF's under the doitintl-cmp-bq dataset, as well as two more derived tables called pegUtilsT2 and queries. These support the FinOps in running over aggregated data, rather than raw data. The following queries are executed as part of the process (2) above:
    Aggregations into pegUtilsT2 and queries tables
    Daily run of cost simulation queries to provide the cost savings information and potential savings.
    Daily run of FinOps calculation queries to provide the richly detailed information you see in the CMP on your BigQuery usage.
View the bite-sized video below for a quick tutorial on the BigQuery FinOps Dashboard.

What does it mean when I see "Not yet allocated" in the Cost Explorer?

You will see "not yet allocated" when analyzing dimensions like Projects in the BigQuery Explorer when we aren't able to detect which table a query is scanning.
This could be because of queries like "SELECT 1" or queries we aren't able to parse due to complexity (parser times out), or it uses a function/statement which we don't support yet.

I understand that BigQuery datasets are created in the background. How much does BigQuery finops cost monthly?

Typically, it’s less than $10 per month.
A screenshot highlighting the Not yet allocated figure
Last modified 6d ago