ADA-C01 Question 392
Single answerMonitor and calculate data transfer costs and replication costsA Snowflake account in AWS us-east-1 replicates a 12 TB production database to a secondary account in AWS eu-west-1 for disaster recovery. The DBA wants to estimate ongoing charges and identify where to monitor them after enabling replication and regular refreshes. Which statement best describes how Snowflake costs should be calculated and monitored for this setup?
- A
Only warehouse credits used to query replicated data are charged by Snowflake; the cross-region movement itself is included at no additional cost and can be monitored only through cloud provider billing.
- B
Replication incurs Snowflake charges for the data transferred during refresh operations and for the storage consumed by the replicated database in the target account; these charges can be monitored in Snowflake usage views such as data transfer and replication-related account usage views.
- C
Replication charges are based on the full logical size of the primary database every time a refresh runs, even if only a small subset of micro-partitions changed; monitoring requires querying INFORMATION_SCHEMA only.
- D
Replication between regions is billed only once when the secondary database is initially created; subsequent refreshes do not incur transfer charges because Snowflake sends metadata-only deltas.
Show answer and explanation
Correct answer: B
Explanation
For cross-region database replication, administrators should think about two main Snowflake cost components: 1) data transfer generated by replication/refresh activity, especially when data is copied across regions, and 2) storage consumed by the replicated copy in the secondary account/region. In practice, the initial replication may move a large volume of data, while subsequent refresh costs depend on the amount of changed data that must be synchronized. To monitor these costs, Snowflake provides Account Usage views for billing/usage analysis, including views related to data transfer and replication behavior, which are more appropriate for historical monitoring than relying on INFORMATION_SCHEMA alone. This aligns with Snowflake guidance to use organization/account usage data to analyze billing drivers and to distinguish storage, compute, and transfer-related charges when planning disaster recovery architectures.
- A. Incorrect.
Incorrect. Snowflake replication across regions/cloud regions can incur Snowflake data transfer charges in addition to any compute used for queries. It is a common misconception that only warehouses matter. While cloud infrastructure underlies the service, administrators should use Snowflake-provided usage and billing views to monitor Snowflake-attributed transfer and replication costs rather than relying only on external cloud bills.
- B. Correct.
Correct. In a cross-region replication scenario, Snowflake can charge for data transfer associated with replication/refresh activity, and the replicated objects also consume storage in the target account, which contributes to cost. Administrators typically monitor these charges using Snowflake Account Usage views related to data transfer and replication/storage consumption. This reflects the practical model: estimate transfer costs from replicated changed data over time, then add the storage footprint maintained in the secondary location.
- C. Incorrect.
Incorrect. Snowflake replication is not billed as if the entire database were recopied on every refresh regardless of change volume. A practical estimate is based on ongoing changed data that must be synchronized, not repeatedly charging for the full logical database size each cycle. Also, limiting monitoring to INFORMATION_SCHEMA is incorrect; Account Usage views are the primary source for historical cost and usage analysis.
- D. Incorrect.
Incorrect. Initial synchronization is not the only billable event. Subsequent refreshes can continue to generate transfer charges when data changes must be synchronized to the secondary database. The idea that all later refreshes are metadata-only and therefore free is a misleading simplification and does not reflect how ongoing replication cost should be estimated.