ADA-C01 SnowPro Advanced Administrator Question and Answers

Option C creates a network policy that allows only the IP address 192.168.1.100 and blocks all other IP addresses using the CIDR notation 0.0.0.0/01. It then applies the network policy to the user ABC, who has the ACCOUNTADMIN role. This ensures that only this user can access the Snowflake account from the specified IP address, while blocking their access from anywhere else. Option A does not block any other IP addresses, option B applies the network policy to the role instead of the user, and option D uses an invalid CIDR notation.

The REFRESH DATABASE command is used to refresh a secondary database with the latest data and metadata from the primary database1. The command must be executed in the target account where the secondary database resides2. Therefore, the answer is A, as the location is AWS_WEST and the command is REFRESH DATABASE PROD_AWS_WEST REFRESH. The other options are incorrect because they either use the wrong location, the wrong command, or the wrong database name.

According to the Snowflake Warehouse Cost Optimization blog post, one of the strategies to reduce the cost of running a warehouse is to use a multi-cluster warehouse with auto-scaling enabled. This allows the warehouse to automatically adjust the number of clusters based on the concurrency demand and the queue size. A multi-cluster warehouse can also be configured with a minimum and maximum number of clusters, as well as a scaling policy to control the scaling behavior. This way, the warehouse can handle the parallel load queries efficiently without wasting resources or credits. The blog post also suggests using a smaller warehouse size, such as SMALL or XSMALL, for loading data, as it can perform better than a larger warehouse size for small INSERTs. Therefore, the best option to reduce the costs while minimizing the overall load times for migrating data warehouse history is to keep the warehouse as a SMALL or XSMALL and configure it as a multi-cluster warehouse to handle the parallel load queries. The other options are incorrect because:

•A. Deploying another 2XL warehouse to handle a portion of the load queries will not reduce the costs, but increase them. It will also introduce complexity and potential inconsistency in managing the data loading process across multiple warehouses.

•B. Changing the 2XL warehouse to 4XL will not reduce the costs, but increase them. It will also provide more compute resources than needed for small INSERTs, which are not CPU-intensive but I/O-intensive.

•D. Converting the INSERTs to several tables will not reduce the costs, but increase them. It will also create unnecessary data duplication and fragmentation, which will affect the query performance and data quality.

According to the Snowflake documentation1, secure views are only exposed to authorized users who have been granted the role that owns the view. Therefore, applying RBAC directly onto the partner’s shared secure views (option A) is not possible, as the administrator does not own those views. Individually granting imported privileges onto the schema in the share (option B) is also not feasible, as the privileges granted on the schema do not apply to existing secure views, only to future ones2. Cloning the data and inserting it into a company-owned share (option C) is not recommended, as it would create unnecessary duplication of data and increase storage costs. The best option is to create views over the incoming shared database and apply the desired RBAC onto these views (option D). This way, the administrator can control the access to the data based on the roles in their account, without modifying the original data or views from the partner.

According to the Snowflake documentation1, data sharing is a feature that allows sharing selected objects in a database in one account with other accounts in the same organization, without copying or transferring any data. Data sharing is supported across regions and across cloud platforms, but it requires enabling account database replication for both the source and target accounts2. Data replication is a feature that allows replicating objects from a source account to one or more target accounts in the same organization, providing read-only access for the replicated objects. Data replication is also supported across regions and across cloud platforms, but it incurs additional storage costs for the replicated data2. Therefore, the best way to share data bi-directionally among all offices, while minimizing costs, is to use data sharing among offices in the same region, which does not require replication or additional storage, and use replication among offices across the continents, which provides near real-time access to the shared data. Option A is incorrect because using data replication everywhere would increase the costs associated with additional storage and compute resources for the replicated data. Option B is incorrect because using the PUT command to move files to an Amazon S3 bucket and Azure Blobs, and using an external file management application to move files within the corporate VPC, would not leverage the benefits of Snowflake’s data sharing and replication features, and would also incur additional costs and complexity for data transfer and synchronization. Option C is incorrect because moving all the Snowflake accounts to a single region would violate the PII regulations in some of the European countries, and would also incur additional costs and complexity for data migration and consolidation.

According to the AT | BEFORE documentation, the AT or BEFORE clause is used for Snowflake Time Travel, which allows you to query historical data from a table based on a specific point in the past. The clause can use one of the following parameters to pinpoint the exact historical data you wish to access:

•TIMESTAMP: Specifies an exact date and time to use for Time Travel.

•OFFSET: Specifies the difference in seconds from the current time to use for Time Travel.

•STATEMENT: Specifies the query ID of a statement to use as the reference point for Time Travel.

Therefore, the queries that will allow the user to view the historical data that was in the table before the query was executed are:

•B. SELECT * FROM my_table AT (TIMESTAMP => ‘2021-01-01 07:00:00’ :: timestamp); This query uses the TIMESTAMP parameter to specify a point in time that is before the query execution time of 07:01.

•D. SELECT * FROM my table PRIOR TO STATEMENT ‘8e5d0ca9-005e-44e6-b858-a8f5b37c5726’; This query uses the PRIOR TO STATEMENT keyword and the STATEMENT parameter to specify a point in time that is immediately preceding the query execution time of 07:01.

•F. SELECT * FROM my_table BEFORE (STATEMENT => ‘8e5d0ca9-005e-44e6-b858-a8f5b37c5726’); This query uses the BEFORE keyword and the STATEMENT parameter to specify a point in time that is immediately preceding the query execution time of 07:01.

The other queries are incorrect because:

•A. SELECT * FROM my table WITH TIME_TRAVEL (OFFSET => -60*30); This query uses the OFFSET parameter to specify a point in time that is 30 minutes before the current time, which is 07:30. This is after the query execution time of 07:01, so it will not show the historical data before the query was executed.

•C. SELECT * FROM TIME_TRAVEL (‘MY_TABLE’, 2021-01-01 07:00:00); This query is not valid syntax for Time Travel. The TIME_TRAVEL function does not exist in Snowflake. The correct syntax is to use the AT or BEFORE clause after the table name in the FROM clause.

•E. SELECT * FROM my_table AT (OFFSET => -60*30); This query uses the AT keyword and the OFFSET parameter to specify a point in time that is 30 minutes before the current time, which is 07:30. This is equal to the query execution time of 07:01, so it will not show the historical data before the query was executed. The AT keyword specifies that the request is inclusive of any changes made by a statement or transaction with timestamp equal to the specified parameter. To exclude the changes made by the query, the BEFORE keyword should be used instead.

According to the Network Policies documentation, a network policy can be applied to an account, a security integration, or a user. If there are network policies applied to more than one of these, the most specific network policy overrides more general network policies. The following summarizes the order of precedence:

•Account: Network policies applied to an account are the most general network policies. They are overridden by network policies applied to a security integration or user.

•Security Integration: Network policies applied to a security integration override network policies applied to the account, but are overridden by a network policy applied to a user.

•User: Network policies applied to a user are the most specific network policies. They override both accounts and security integrations.

Therefore, if both the account_level and user_level network policies are defined, the user_level policy will take effect and the account_level policy will be ignored. The other options are incorrect because:

•The account_level policy will not override the user_level policy, as explained above.

•The user_level network policies will be supported, as they are part of the network policy feature.

•A network policy error will not be generated, as there is no conflict between the account_level and user_level network policies.

According to the Snowflake documentation1, a multi-cluster warehouse is a virtual warehouse that consists of multiple clusters of compute resources that can scale up or down automatically to handle the concurrency and performance needs of the queries submitted to the warehouse. A multi-cluster warehouse has a minimum and maximum number of clusters that can be specified by the administrator. Option A is a possible solution to address the issues happening once a month, as it allows the administrator to use a task to increase the cluster size for the time period that the more complex queries are running and another task to reduce the size of the cluster once the complex queries complete. This way, the warehouse can have more resources available to handle the complex queries without reaching the maximum limit of 10 clusters, and then return to the normal cluster size to save costs. Option B is another possible solution to address the issues happening once a month, as it allows the administrator to have the group running the complex monthly queries use a separate appropriately-sized warehouse to support their workload. This way, the warehouse can isolate the complex queries from the standard workload and avoid queue time and resource contention. Option C is not a recommended solution to address the issues happening once a month, as it would increase the costs and complexity of managing the multi-cluster warehouse, and may not solve the underlying problem of inefficient queries. Option D is a good practice to improve the performance of the queries, but it is not a direct solution to address the issues happening once a month, as it requires analyzing and optimizing the complex queries using clustering keys or materialized views, which may not be feasible or effective in all cases. Option E is not a recommended solution to address the issues happening once a month, as it would increase the costs and waste resources by starting more clusters than needed for the standard workload.

This command will generate a sample of 20% of the table by using the SYSTEM (or BLOCK) sampling method, which selects each block of rows with a probability of 20/100. This method is suitable for taking data by blocks of rows, as the question requires. According to the Snowflake documentation, “SYSTEM (or BLOCK): Includes each block of rows with a probability of p/100. Similar to flipping a weighted coin for each block of rows. This method does not support fixed-size sampling.” The other options are either incorrect or do not meet the requirements of the question. Option A uses the BERNOULLI (or ROW) sampling method, which selects each row with a probability of 20/100, but does not take data by blocks of rows. Option C uses the BLOCK sampling method, but specifies a fixed number of rows (20) instead of a percentage (20%). Option D uses the SYSTEM sampling method, but specifies a seed value (1), which makes the sampling repeatable, contrary to the question.

According to the Snowflake documentation1, the AUTO_SUSPEND property specifies the number of seconds of inactivity after which a warehouse is automatically suspended. If the property is set to 0 or NULL, the warehouse never suspends automatically. For a warehouse that supports several dashboards, issuing various queries on the same database, setting AUTO_SUSPEND to 0 or NULL can help to keep the data cache warm, which means that the data used by the queries is already loaded into the warehouse memory and does not need to be fetched from the storage layer. This can improve the performance of similar queries that access the same data. Option A is incorrect because setting AUTO_SUSPEND to 0 or NULL does not save costs on warehouse shutdowns and startups, but rather increases the costs by keeping the warehouse running continuously. Option B is incorrect because setting AUTO_SUSPEND to 0 or NULL does not run the warehouse as little as possible, but rather runs the warehouse as much as possible. Option D is incorrect because setting AUTO_SUSPEND to 0 or NULL does not affect the query result cache, which is a separate cache that stores the results of previous queries for a period of time. The query result cache is not dependent on the warehouse state, but on the query criteria2.

Option D is the correct answer because it follows the steps described in the Snowflake documentation for importing data from the Snowflake Marketplace. The ACCOUNTADMIN role needs to grant the IMPORT SHARE privilege on the account to the DP_TEAM role, which allows the role to import data from any provider in the marketplace. The ACCOUNTADMIN role also needs to grant the CREATE DATABASE privilege on the account to the DP_TEAM role, which allows the role to create a database from a share. Option A is incorrect because there is no MARKETPLACE privilege in Snowflake. Option B is incorrect because the USAGE privilege on SNOWFLAKE_MARKETPLACE is not sufficient to import data from the marketplace. Option C is incorrect because there is no IMPORTED PRIVILEGES privilege in Snowflake.

According to the Snowflake documentation1, the MONITOR privilege on a warehouse grants the ability to view the warehouse usage and performance metrics, such as the number of credits consumed, the average and maximum run time, and the number of queries executed. The USAGE privilege on a warehouse grants the ability to use the warehouse to execute queries and load data. Therefore, the minimal privileges on the warehouse that should be granted to the role to meet the requirements for the application are MONITOR and USAGE. Option A is incorrect because the OPERATE privilege on a warehouse grants the ability to start, stop, resume, and suspend the warehouse, which is not required for the application. Option B is incorrect because the MODIFY privilege on a warehouse grants the ability to alter the warehouse properties, such as the size, auto-suspend, and auto-resume settings, which is not required for the application. Option E is incorrect because the OWNERSHIP privilege on a warehouse grants the ability to drop the warehouse, grant or revoke privileges on the warehouse, and transfer the ownership to another role, which is not required for the application.

According to the Snowflake documentation1, when an object in a schema is cloned, any future grants defined for this object type in the schema are applied to the cloned object unless the COPY GRANTS option is specified in the CREATE statement for the clone operation. However, some objects may still reference the source object or external objects after cloning, which may cause issues with access or functionality. These objects include:

•Sequences: If a table column references a sequence that generates default values, the cloned table may reference the source or cloned sequence, depending on where the sequence is defined. To avoid conflicts, the sequence reference should be re-pointed to the desired sequence using the ALTER TABLE command2.

•Storage integrations: If a stage or a table references a storage integration, the cloned object may still reference the source storage integration, which may not be accessible or valid in the new environment. To avoid errors, the storage integration reference should be re-pointed to the desired storage integration using the ALTER STAGE or ALTER TABLE command34.

•Views, secure views, and materialized views: If a view references another view or table, the cloned view may still reference the source object, which may not be accessible or valid in the new environment. To avoid errors, the view reference should be re-pointed to the desired object using the CREATE OR REPLACE VIEW command5.

1: Cloning Considerations | Snowflake Documentation 2: [ALTER TABLE | Snowflake Documentation] 3: [ALTER STAGE | Snowflake Documentation] 4: [ALTER TABLE | Snowflake Documentation] 5: [CREATE VIEW | Snowflake Documentation]

The EXPLAIN command returns the logical execution plan for a query, which shows the upper bound estimates for the number of partitions and bytes that might be scanned by the query1. However, these estimates do not account for the runtime optimizations that Snowflake performs to improve the query performance and reduce the resource consumption2. One of these optimizations is join pruning, which eliminates unnecessary partitions from the join inputs based on the join predicates2. This can result in fewer partitions and bytes scanned than the estimates from the EXPLAIN output3. Therefore, the actual partitions scanned value in the Query Profile can be lower than the partitionsAssigned value in the EXPLAIN output4.

To configure a Snowflake account to use AWS PrivateLink, the Administrator needs to create CNAME records in the DNS that point to the private endpoints provided by Snowflake. This allows the clients to connect to Snowflake using the same URL as before, but with private connectivity. According to the Snowflake documentation, “After you have created the VPC endpoints, Snowflake provides you with a list of private endpoints for your account. You must create CNAME records in your DNS that point to these private endpoints. The CNAME records must use the same hostnames as the original Snowflake URLs for your account.” The other options are either incorrect or not sufficient to configure AWS PrivateLink. Option B is not necessary, as the Administrator can enable AWS PrivateLink using the SYSTEM$AUTHORIZE_PRIVATELINK function1. Option C is not recommended, as it may prevent some data traffic from reaching Snowflake, such as large result sets stored on AWS S32. Option D is not related to AWS PrivateLink, but to Snowflake Connectivity Diagnostic (SnowCD), which is a tool for diagnosing network issues between clients and Snowflake3.

According to the Understanding & Using Time Travel documentation, Time Travel is a feature that allows you to query, clone, and restore historical data in tables, schemas, and databases for up to 90 days. To meet the requirements of the scenario, the following best practices should be followed:

•The fact and dimension tables should have the same DATA_RETENTION_TIME_IN_DAYS. This parameter specifies the number of days for which the historical data is preserved and can be accessed by Time Travel. To ensure that the fact and dimension tables can be reverted to a consistent state in case of any anomalies in the latest load, they should have the same retention period. Otherwise, some tables may lose their historical data before others, resulting in data inconsistency and quality issues.

•The fact and dimension tables should be cloned together using the same Time Travel options to reduce potential referential integrity issues with the restored data. Cloning is a way of creating a copy of an object (table, schema, or database) at a specific point in time using Time Travel. To ensure that the fact and dimension tables are cloned with the same data set, they should be cloned together using the same AT or BEFORE clause. This will avoid any referential integrity issues that may arise from cloning tables at different points in time.

The other options are incorrect because:

•Related data should not be placed together in the same schema. Facts and dimension tables should each have their own schemas. This is not a best practice for Time Travel, as it does not affect the ability to query, clone, or restore historical data. However, it may be a good practice for data modeling and organization, depending on the use case and design principles.

•The DATA_RETENTION_TIME_IN_DAYS should be kept at the account level and never used for lower level containers (databases and schemas). This is not a best practice for Time Travel, as it limits the flexibility and granularity of setting the retention period for different objects. The retention period can be set at the account, database, schema, or table level, and the most specific setting overrides the more general ones. This allows for customizing the retention period based on the data needs and characteristics of each object.

•Only TRANSIENT tables should be used to ensure referential integrity between the fact and dimension tables. This is not a best practice for Time Travel, as it does not affect the referential integrity between the tables. Transient tables are tables that do not have a Fail-safe period, which means that they cannot be recovered by Snowflake after the retention period ends. However, they still support Time Travel within the retention period, and can be queried, cloned, and restored like permanent tables. The choice of table type depends on the data durability and availability requirements, not on the referential integrity.

According to the Snowflake documentation1, a multi-cluster warehouse is a virtual warehouse that consists of multiple clusters of compute resources that can scale up or down automatically to handle the concurrency and performance needs of the queries submitted to the warehouse. A multi-cluster warehouse has a minimum and maximum number of clusters that can be specified by the administrator. Option B is the most efficient way to support the group of users, as it allows the administrator to create a multi-cluster warehouse with MIN_CLUSTERS set to 2, which means that the warehouse will always have two clusters running to handle the standard workload. The warehouse can also auto-scale up to the maximum number of clusters (which can be set according to the peak workload) when there is a spike in demand, and then scale down when the demand decreases. The warehouse can also auto-resume and auto-suspend, which means that the warehouse will automatically start when a query is submitted and automatically stop after a period of inactivity. The administrator can also give USAGE privileges to the ANALYST role, which means that the users can use the warehouse to execute queries and load data, but not modify or operate the warehouse. Option A is not efficient, as it requires the users to manually start and stop the warehouse, and increase the number of clusters as needed, which can be time-consuming and error-prone. Option C is not efficient, as it creates a standard X-Large warehouse, which is equivalent to four small clusters, which may be more than needed for the standard workload, and may not be enough for the peak workload. Option D is not efficient, as it creates four virtual warehouses of different sizes, which can be confusing and cumbersome for the users to select the appropriate warehouse based on how many queries are being run, and may also result in wasted resources and costs.

The user needs the OPERATE privilege on the task to suspend and resume it, and the USAGE privilege on the database and schema containing the task to access it1. The EXECUTE TASK privilege is not required for suspending and resuming a task, only for triggering a new run1. The OWNERSHIP privilege on the task or the database and schema would allow the user to modify or drop the task, which is not desired.