Choosing a Snowflake Data Sharing Method: Listings, Marketplace, Exchanges, Clean Rooms

1.Four sharing methods and what sets them apart

Every Snowflake sharing method packages database objects into a share that consumer accounts query read-only. The documentation names four methods: a listing, a direct share, a Data Exchange and a clean room. They differ in reach and control, and the table and the matching check below pin down each one.

So a direct share fits a same-region partner where you need nothing extra. Choose a listing when you need to cross regions, publish on the Marketplace, charge for access, add descriptions and sample queries, or see usage metrics. A listing can be offered privately to specific accounts or publicly on the Marketplace, and an existing direct share can be converted into a listing. With listings, direct shares and exchanges, the consumer queries the data directly. Choose a clean room when the consumer must not have that unrestricted access.

Sources12

2.Sharing across regions and cloud providers

There are two routes. The manual route uses replication. It works for accounts on AWS, Google Cloud Platform and Azure, so the same procedure covers a different region and a different cloud provider. You first create an account in the target region and link it to your organization. An organization administrator enables replication for the source and target accounts. You then create a replication group that holds the database and the share:

USE ROLE ACCOUNTADMIN;

CREATE REPLICATION GROUP my_rg
  OBJECT_TYPES = databases, shares
  ALLOWED_DATABASES = db1
  ALLOWED_SHARES = share1
  ALLOWED_ACCOUNTS = acme.account_2;

In the target account, you create a secondary group AS REPLICA OF the primary, refresh it, and then add consumer accounts that are local to that region to the replicated share. Setting REPLICATION_SCHEDULE on the primary group automates the refreshes. Plan for four things. You need only one copy per region, not one per consumer. A view that references several databases needs all of them in the group. Replication copies whole databases, so to cut cost you can copy only the needed rows into a new database with streams and tasks. And you must confirm that no legal or regulatory rule forbids moving the data to that location.

The second route, Cross-Cloud Auto-Fulfillment, works for listings, whether they go to specific consumers or onto the Marketplace. Snowflake replicates the listing's data product, meaning the share or application package attached to it, to consumer regions as they need it. You don't run replication or approve requests yourself. You can turn it on while the listing is in draft or after it is published. It isn't available on trial accounts.

The limits on this route are the ones exams test. Which route a listing may use depends on how it is offered. A listing shared with specific consumer accounts is a private listing: Snowsight detects when the target account is in another region and enables auto-fulfillment for you, and you cannot replicate a private listing manually. Paid listings must use auto-fulfillment. Free or limited trial listings on the Marketplace may use auto-fulfillment or manual replication. Application packages use auto-fulfillment. Timing follows demand: a private listing is fulfilled after the specified consumers get it, and a public Marketplace listing after a consumer in that region gets it. Making a data product available in other regions costs the provider extra. Snowflake manages a secure share area (SSA) that receives the data product in each region, one per organization in each commercial region. Virtual Private Snowflake environments and government regions each get a separate SSA.

Sources34

3.Snowflake Marketplace versus a Data Exchange

Snowflake Marketplace is the public route. A provider publishes listings there to share curated data with many consumers at once, instead of managing a sharing relationship with each one. Paid listings let it charge. It can also publish free datasets to generate interest, or samples that it provides on request or customizes. Marketplace is available on AWS, Google Cloud and Azure, except Microsoft Azure Government. Users who reach it over private connectivity need a CNAME record first. Listings offered privately, through an exchange or on the Marketplace come with consumer usage metrics.

A Data Exchange is the private route. It is a data hub for a group of members you invite, such as internal departments, suppliers or partners. You request it from Snowflake: it must be provisioned for your account. The account that hosts it is the Data Exchange Admin. Its ACCOUNTADMIN adds and removes members, marks each as a provider, a consumer or both, and can delegate those privileges to other roles. Providers publish free or personalized listings. Consumers browse them, and getting data needs ACCOUNTADMIN or IMPORT SHARE. Usage data for exchange listings appears only in the Data Sharing Usage schema views.

To choose between a Data Exchange and a private listing, look at the audience. The documentation positions a Data Exchange for a specific group of consistent business partners, and listings, offered to specific consumers or publicly, for a variety of consumers inside and outside your organization. A listing offered privately to specific accounts is the first choice, and a Data Exchange is the alternative when that isn't an option. The exchange adds a managed group in which the admin controls who can publish and consume. Marketplace listings come in several types: free, paid, personalized (consumers request access) and limited trial (instant access for 1 to 90 days). Providers can't convert a listing from one type to another.

Sources5672

4.Sharing with consumers who have no Snowflake account

The classic answer is a reader account, which the provider creates as a managed account. The reader account belongs to the provider, which pays for it, and it can consume data only from that provider. Its users can query imported data, but they cannot load, insert, update or do any other DML.

CREATE MANAGED ACCOUNT <name>
    ADMIN_NAME = <username> , ADMIN_PASSWORD = <user_password> ,
    TYPE = READER ,
    [ COMMENT = '<string_literal>' ]

A newer option, in Public Preview, serves consumers who use no Snowflake at all. They query live Iceberg tables through the standard Iceberg REST Catalog APIs from their own tools. The provider creates an external consumer, a restricted identity tied to a region, and gives it a programmatic access token (PAT). The PAT secret must be saved when it is issued, because it cannot be retrieved later. The provider then attaches the share to that consumer through an external listing. This needs ACCOUNTADMIN or the CREATE EXTERNAL CONSUMER privilege. If the account uses a network policy, each external consumer needs an entry in it. During the preview, the data can be shared only in the provider's own region.

CREATE EXTERNAL CONSUMER test_ext_consumer;

SHOW EXTERNAL CONSUMERS LIKE '%test_ext_consumer%';

The two options differ in what the consumer gets. A reader account is a Snowflake account that the provider creates, in the provider's region and on the provider's edition. Its users log in, create warehouses and query the share with SQL, and the provider pays for all of it, so a resource monitor is wise. Open sharing needs no Snowflake account at all: the consumer connects with an Iceberg REST Catalog client and a PAT, and the shared data is Iceberg tables. When a question asks how to share with someone who has no Snowflake account, the documented established answer is a reader account. Open sharing is a newer alternative that the documentation marks Public Preview. During the preview it supports only PAT authentication and only the provider's own region, so treat it as a limited option and not as a replacement for reader accounts.

Sources189

5.Snowflake Data Clean Rooms: sharing with controlled analysis

Every method so far gives the consumer direct access to the shared data. A Snowflake Data Clean Room is for collaborations where that would expose too much. Collaborators share data, and it can be queried only through templates added to the collaboration. Collaborators review new templates and data sources, and approve or reject them. All analysis runs inside the clean room. Participants get back aggregated results and insights, never the raw rows, and the collaborator sharing the data decides which analyses the others may run. The documentation lists the benefits as stronger privacy, deeper insight from combining multiple sources, less risk of unauthorized access, and private machine learning on the combined data.

Each collaborator holds one or more roles. The Owner creates the collaboration and decides who has which role. A Data Provider supplies data and sets the policies on it. An Analysis Runner runs the templates on the data it is permitted to use. Templates are the JinjaSQL queries that do the analysis, and any collaborator can submit one for the others to approve. Results can be returned directly or activated, meaning exported to an approved destination such as a collaborator's own Snowflake account. Edition matters: data providers must use Enterprise Edition, owners and analysis runners can use Standard Edition, and activating results to another Snowflake account also needs Enterprise Edition. A provider can invite non-Snowflake users to a clean room as well as Snowflake accounts.

Sources21011

Exam traps

Each one states something that sounds right. Open it to see what is actually true.

1. 1.A direct share can reach a consumer in another region or on another cloud if you add that consumer's account to it.Why is that wrong?

   Direct shares work only within one region. Other regions and clouds need replication groups or a listing with Cross-Cloud Auto-Fulfillment.

   Covered in Sharing across regions and cloud providers

2. 2.A reader account works like a small full account, so the consumer can load their own data into it to enrich the shared data.Why is that wrong?

   Reader account users can only query imported data. Loading, inserting, updating and other DML are not allowed.

   Covered in Sharing with consumers who have no Snowflake account