Practice Test 4 | Google Cloud Certified Professional Data Engineer | Dumps | Mock Test
You are working the social networking company having billions of customers. Data scientist of your company has created machine learning algorithm using Big Query to predict a certain score for each user based on their social profile. These scores will be used by web application per individual basis. The web application will need a score for each user under 100ms. Result from the model can be fetched by running a query using ML. Predict. How should the application get the score in cost efficient manner?
A. Create authorized view in separate dataset, which will run the query containing ML.Predict. This query will have WHERE clause supplying user_id to fetch the score for that particular user.
B. Create separate table having only user_id and score. Application can run the query using service account.
C. Run the query for all the users to fetch out results. Store the result in BigTable having row_key on user_id. Provide only read access to the application to fetch out the results.
D. Run the query with ML.Predict and export the result in AVRO format to cloud storage. Create RestAPI using compute engine to read the results from this cloud storage bucket.
The correct answer is C
Option A is incorrect. Creating authorized view and supplying user_id each time will run the Model again. This will incur charges and will not be cost efficient. As data is for billions of customer, running prediction query each time is not appropriate solution.
Option B is incorrect. As table contains billions of customer, running query against this table will scan the table everytime resulting in using Big Query slots and will not be cost effective. Also, query time might not under 100ms.
Option C is correct. BigTable can stores billions of records and fetching from BigTable is fast if row_key is correctly designed. In this case, storing all the prediction result in BigTable and then giving only read access to the web application would be the correct approach. As BigTable has read performance of 6ms, web application can get the score of each user under 100ms.
Option D is incorrect. As we need the result under 100ms, running a RestApi to fetch the result will take more time. Data could be in TB/PB as billions of users are registered on social networking site.
Therefore, this would not be appropriate solution.
Comments are closed, but trackbacks and pingbacks are open.