More on Data Sharing
- ✅ Granting Access to Consumer Accounts
- Share data securely with existing Snowflake accounts using account-level privileges.
- Consumers query live shared data without data duplication.
- Consumers use their own compute resources, reducing provider’s cost.
- ✅ Creating Reader Accounts
-
For consumers without Snowflake, create Managed Reader Accounts.
-
Reader Accounts:
- Fully managed by the provider.
- Isolated, secure environment for the consumer.
- Provider controls billing and administration.
- ❓ Why use Reader Accounts instead of just sharing directly
- ✅ Consumer doesn’t have Snowflake.
- ✅ Provider needs full control over security and billing.
- ✅ Avoids relinquishing admin rights to the consumer. (Direct sharing is for existing Snowflake accounts where the consumer handles billing and access control.)
- ⚠️ Limitations and Control Considerations
⚠️ 4. Limitations & Control Considerations
-
Shared data is read-only:
- ❌ Cannot modify or delete data.
- ❌ Cannot insert new rows.
- ❌ Cannot re-share data.
- Governance remains with the provider using
IMPORTED PRIVILEGES.
- 🔁 Can Consumers Modify Data or Share it Further?
-
No!
-
Data integrity remains with the provider:
- Read-only access.
- Cannot alter or redistribute data.
- 🧠 Best Practices for Managing Shared Access
✔ Follow governance principles with roles, masking, auditing, and periodic access reviews. ✔ Use Secure Views for masking sensitive data. ✔ Keep separate schemas for shared data. ✔ Monitor access history for audits. ✔ Schedule regular privilege reviews. ✔ Tag shares with labels/comments for governance.
💰 Billing
-
Simply,
-
Provider pays for storage & compute in Reader Accounts.
-
In direct sharing, consumer pays for their compute usage.
Billing Methods for Different Types of Data Sharing in Snowflake
| Sharing Type | Billing Responsibility | Details and Other Aspects |
|---|---|---|
| Internal Sharing | No additional billing; uses same account | Data shared within the same Snowflake account or org. Since it does not consume extra compute or storage beyond the existing account’s resources, no separate billing applies. |
| External Sharing (Account-Level Sharing) | Consumer pays for compute | Provider stores data and pays for storage costs. The consumer queries shared data using their own compute (virtual warehouses) and pays for that compute usage. Storage billing remains with the provider. |
| Reader Account Sharing | Provider pays for both storage and compute | Provider creates and manages Reader Accounts for consumers without Snowflake accounts. Here, the provider incurs costs for both the storage of shared data and the compute resources used by Reader Accounts to query data. |
Additional Billing Aspects
- Storage Costs: Always billed to the provider who owns the data, regardless of the sharing method.
- Compute Costs: Vary by sharing type
- For external consumers with their own Snowflake accounts, compute is billed to the consumer.
- For Reader Accounts, compute is billed to the provider, since the provider manages those accounts.
- Data Transfer Costs: Minimal within Snowflake but provider might incur costs depending on network egress in cross-region or cross-cloud shares.
- Billing Control:
- Reader Accounts give the provider full control over usage and associated billing.
- External account sharing delegates compute and billing responsibility to the consumer.
❓ Why Not Just Share Directly?
| Situation | Recommended Approach | Why |
|---|---|---|
| Consumer already has a Snowflake account | Use Direct Account Sharing (aka External Sharing) | Because the consumer can receive the data in their own Snowflake account without extra setup. |
| Consumer does NOT have a Snowflake account | Use Reader Account | Because you can create and manage a Snowflake account on their behalf so they can access data. |
| You (Provider) want FULL control over consumer's access | Reader Account | Reader accounts are fully managed by the provider, including users, roles, billing (consumer cannot change these). |
| You want the consumer to manage their own users and roles | Direct Account Sharing | Because in direct sharing, the consumer controls access inside their account. |
⚠️ Limitations & Control Considerations
❌ Can Consumers Edit or Modify Shared Data?
No.
- Shared data is read-only.
- Consumers cannot modify, insert, or delete rows from the shared database.
- They cannot grant access to other users/accounts.
- The
IMPORTED PRIVILEGESclause allows internal roles in consumer accounts to read shared data but not share it further.
🔐 Security Controls in Place
| Control Point | Provider Responsibility |
|---|---|
| Data Access Permissions | SELECT, USAGE on shared objects |
| Metadata Visibility | Only what’s included in the share |
| Row/Column Access Control | Use Row Access Policies, Dynamic Masking |
| Compute Isolation | Consumer uses own virtual warehouse (except in reader accounts) |
🔁 Can Shared Access Be Revoked or Modified?
Yes. You can:
- ✅ Add or remove objects from the share:
ALTER SHARE share_name ADD TABLE schema.table_name;
- ✅ Revoke access from a consumer:
ALTER SHARE share_name REMOVE ACCOUNTS = ('consumer_account_locator');
- ✅ Drop the share altogether:
DROP SHARE share_name;
🧠 Best Practices for Sharing Access
| Practice | Benefit |
|---|---|
| Use Secure Views | Prevent exposure of sensitive columns |
| Create Separate Shared Schemas | Isolate shared data from production data |
| Monitor Access History | Track who accessed what, when |
| Schedule Reviews of Shared Grants | Maintain governance and compliance |
| Label shares with Tags or Comments | For auditing and documentation |
📌 Summary
| Topic | Key Points |
|---|---|
| Grant Access to Consumer | Used when the recipient has a Snowflake account |
| Create Reader Account | Used for external consumers without Snowflake |
| Read-Only Access | Consumers cannot edit, export, or re-share the data |
| Billing | Reader Account: Provider pays; Regular Consumer: Consumer pays |
| Modification/Revoke | Provider can manage and revoke access at any time |
| Security | Strong RBAC, masking, row-level security, and audit capabilities |