Task 4.3: Sharing data.

• ADirect IAM grants of Data Viewer to each customer's users on the datasets

  Incorrect

  IAM grants work across organizations but give no catalog or self-service subscription, and 300 customers' identities would be managed grant by grant.

• BA nightly table copy into a dataset in each customer's project

  Incorrect

  Copies duplicate the data 300 times and leave customers a day behind.

• CA BigQuery sharing exchange with listings that customers subscribe to

  Correct

  Subscribing to a listing creates a read-only linked dataset that points at the shared data without copying it, and subscribers pay for their own queries.

• DExports of the tables to Parquet in buckets that each customer can read

  Incorrect

  Exports are copies that go stale, and customers would query files instead of live tables.

BigQuery sharing, formerly Analytics Hub, lets publishers put datasets into exchanges as listings. Subscribers discover listings and subscribe, which creates a linked dataset in their project; queries are billed to the subscriber, storage to the publisher, and nothing is copied.

• AA BigQuery sharing listing of the retailer's customer table

  Incorrect

  An ordinary listing lets subscribers query the shared rows themselves, which the two parties forbid.

• BAuthorized views that the retailer grants to the brand's analysts

  Incorrect

  An authorized view can hide columns or rows, but it does not force every query to aggregate over groups of a minimum size.

• CA shared dataset with column-level policy tags on the email column

  Incorrect

  Policy tags restrict who can read a column; they do not limit queries to aggregates.

• DA BigQuery data clean room with an aggregation threshold analysis rule

  Correct

  Clean rooms let parties analyze shared data without exposing raw rows; an aggregation threshold rule permits only aggregate queries over a minimum number of entities, and the clean room stops copying of the raw data.

Data clean rooms add privacy controls on top of BigQuery sharing. Analysis rules on the shared views, such as the aggregation threshold rule, restrict subscribers to aggregate queries, and the clean room keeps them from copying or exporting the raw data.

• AGrant the partners Data Viewer on the central dataset directly

  Incorrect

  Direct access would expose the curated tables themselves, not just the views.

• BAuthorize the partner-reporting dataset on the central dataset

  Correct

  An authorized dataset gives every current and future view in it access to the shared dataset, and counts as one entry in the access list.

• CReplace the views with materialized views authorized one by one

  Incorrect

  Materialized views still need individual authorization, so the access list keeps growing.

• DCopy the curated tables into the partner-reporting dataset

  Incorrect

  Copies duplicate data, go stale and must be secured separately.

A dataset's access list can hold at most 2,500 authorized resources. When many views need access, authorizing the dataset that contains them gives all of its views, including ones created later, access in one entry.

• AEmail signed URLs to exported files once a buyer has paid

  Incorrect

  Signed URLs hand out file copies and leave billing and discovery to the firm.

• BGrant each buyer's group Data Viewer once it signs a contract

  Incorrect

  Direct grants can work for known customers but offer no marketplace discovery or billing.

• CPublish a commercial listing through Google Cloud Marketplace

  Correct

  Commercial listings can be offered through Cloud Marketplace, where unknown buyers find them, purchase with their Google Cloud billing and then get access to the data.

• DMake the dataset public with allAuthenticatedUsers access

  Incorrect

  Public access gives the data away for free and offers no way to charge.

Selling data to buyers the firm does not know yet needs discovery and billing as well as access. Commercial listings can be offered on Google Cloud Marketplace, so buyers discover the data, pay through Google Cloud billing and query the provider's data in place.

• ASigned URLs for the three objects that expire after 48 hours

  Correct

  A signed URL gives time-limited access to a specific object to anyone who has the URL, with no Google account needed.

• BStorage Object Viewer on the bucket for the auditor's email address

  Incorrect

  An IAM grant needs a Google identity, which the auditor does not have, and does not expire by itself.

• CA BigQuery sharing listing that includes the exported files

  Incorrect

  Sharing listings do not hand out Cloud Storage objects, and their subscribers need Google Cloud identities.

• DRequester Pays on the bucket, with the auditor's billing project

  Incorrect

  Requester Pays changes who pays for access; it grants no access.

Signed URLs carry their own authentication: anyone with the URL can perform the allowed action on the object until it expires, without a Google account. They suit short-lived sharing of specific files with outside parties.

• AA BigQuery scheduled query that emails the results to a group

  Incorrect

  Scheduled queries write results to tables; they do not email reports.

• BA Cloud Run job that renders the report and sends it by email

  Incorrect

  Custom rendering and email code is unnecessary when the report tool delivers PDFs itself.

• CA link to the report pinned in the executives' calendar invite

  Incorrect

  A link still requires opening the report rather than receiving it.

• DA scheduled email delivery of the report each Monday

  Correct

  Data Studio's scheduled delivery emails the report as a PDF attachment with a preview and a link to the full report.

Publishing reports includes pushing them to readers. Data Studio can schedule email delivery of a report as a PDF with a link to the live version, to individual addresses or groups, by its owner or editors.