Exam Certified Data Engineer Professional topic 1 question 35 discussion

Create view. Can't be B as -> without increasing the number of tables that need to be managed

B makes way more sense, the number of tables managed do not increase since the old table won't be used anymore, then the view on top of this table is not another table to manage, just maintains the "original API" of the table to avoid breaking changes in downstream applications

This approach achieves the following key goals: 1. Minimizes Disruption: By creating a view that mirrors the original schema, existing workloads that depend on the current schema remain uninterrupted. Other teams can continue their queries without needing to adjust their logic for the schema change. 2. Meets New Requirements: The new table accommodates the changes required by the customer-facing application, ensuring that the application's updated requirements are fulfilled. 3. Avoids Table Duplication: Instead of maintaining multiple tables for the same dataset, this approach uses a combination of a new table and a view, reducing the overall management burden. 4Flexibility for Future Changes: Views can be adjusted as needed, providing a layer of abstraction. Future schema updates can be handled similarly without directly impacting dependent systems.

Create view -> number of tables stay the same. Option D has overhead

B - but I was wondering between B & D. I do not like D since you replace the table with a view (query costs + you need to change the currently working workflow). Additionally, you create a table that does similar thing to the view - why?

E. not D because, by converting the aggregate table into a view, might introduce performance overheads as every access now potentially involves running complex query logic to reconstruct the desired dataset on-the-fly. This might not be ideal for performance-sensitive applications like business intelligence dashboards.

Option D will increase the Compute cost significantly as all the downstream teams will run the view which has logic for Aggregate table. Option B make more sense with less impact to storage and compute cost which is the original ask for the data engineering team in the question.

I am not sure whether B or D... I believe B increases the number of managed Tables as it states that a CREATE TABLE statement is run before a CREATE VIEW ... the fact that the CREATE VIEW will replace the current table is not really specified... still one could argue that it would be dumb not do it but at this point i would say that D is more precise

The problem with this question is if the order of the sentence matters. B: Create a table then create a view. Teams are interrupted after the creation of the table. D: Create a view then create a table. Teams are not interrupted because they are consuming the view first.

B is Correct. It does not create additional tables. The view mimics the old schema so not to interrupt downstream consumers. It ensures the aggregates are persisted to save on compute. D is incorrect mostly due to the aggregates being baked into the view which is not optimal as each time downstream users query the view the joins and aggregates have to be recomputed.

D will not increate the number of table. It will create a new table and replace the aggregation table with a view. B will create a new table, a new view match old table name and schema, aggregation table still there.

option D is correct docs : https://docs.databricks.com/en/delta/update-schema.html also they specifically says that they dont want to increase managed tables!

B is correct, no new tables, and minimally interrupting other teams in the organization A & E excluded, because they interrupt other teams in the organisation, usually answer that require user communication are wrong answers. C excluded, because it’s used for table creation, not after creation D excluded because it increases the number of tables

B,C and D all state creating a new table, therefore increasing the number of tables to manage. This is exactly what the question says to avoid. "minimally interrupting other teams in the organization without increasing the number of tables that need to be managed" Answer A is the only one that makes sense and is pretty standard operation procedure for databases. E is wrong because you would never update a column comment to inform users of anything.

B is correct. Why not D: Because it will create interruption when you replace the current table with a view and question says minimal interruption

I would go for B. With option B you will run the aggregations once and store in in a table, then present these aggregations in the old schema in a view. With D the aggregations will be done twice, for the old schema view and for the new table.

to me it's b because by creating a new table + the view that will substitute the previous table we still have 1 table. It seems to be the most efficient way to solve this. Not 100% sure though