Exam - 4

Question 1

You work for a global shipping company. You want to train a model on 40 TB of data to predict which ships in each geographic region are likely to cause delivery delays on any given day. The model will be based on multiple attributes collected from multiple sources. Telemetry data, including location in GeoJSON format, will be pulled from each ship and loaded every hour. You want to have a dashboard that shows how many and which ships are likely to cause delays within a region. You want to use a storage solution that has native functionality for prediction and geospatial processing. Which storage solution should you use?
A. BigQuery
B. Cloud Bigtable
C. Cloud Datastore
D. Cloud SQL for PostgreSQL

Answer 1

A. BigQuery

Question 2

You are planning to migrate your current on-premises Apache Hadoop deployment to the cloud. You need to ensure that the deployment is as fault-tolerant and cost-effective as possible for long-running batch jobs. You want to use a managed service. What should you do?
A. Deploy a Cloud Dataproc cluster. Use a standard persistent disk and 50% preemptible workers. Store data in Cloud Storage, and change references in scripts from hdfs:// to gs://
B. Deploy a Cloud Dataproc cluster. Use an SSD persistent disk and 50% preemptible workers. Store data in Cloud Storage, and change references in scripts from hdfs:// to gs://
C. Install Hadoop and Spark on a 10-node Compute Engine instance group with standard instances. Install the Cloud Storage connector, and store the data in Cloud Storage. Change references in scripts from hdfs:// to gs://
D. Install Hadoop and Spark on a 10-node Compute Engine instance group with preemptible instances. Store data in HDFS. Change references in scripts from hdfs:// to gs://

Answer 2

A. Deploy a Cloud Dataproc cluster. Use a standard persistent disk and 50% preemptible workers. Store data in Cloud Storage, and change references in scripts from hdfs:// to gs://

Reason: Dataproc is managed serviced. Standard PD is cheaper than SSD

Question 3

You work for a mid-sized enterprise that needs to move its operational system transaction data from an on-premises database to GCP. The database is about 20
TB in size. Which database should you choose?
A. Cloud SQL
B. Cloud Bigtable
C. Cloud Spanner
D. Cloud Datastore

Answer 3

A. Cloud SQL
Reason: CloudSQL limit 64TB

Question 4

You need to deploy additional dependencies to all of a Cloud Dataproc cluster at startup using an existing initialization action. Company security policies require that Cloud Dataproc nodes do not have access to the Internet so public initialization actions cannot fetch resources. What should you do?
A. Deploy the Cloud SQL Proxy on the Cloud Dataproc master
B. Use an SSH tunnel to give the Cloud Dataproc cluster access to the Internet
C. Copy all dependencies to a Cloud Storage bucket within your VPC security perimeter
D. Use Resource Manager to add the service account used by the Cloud Dataproc cluster to the Network User role

Answer 4

C. Copy all dependencies to a Cloud Storage bucket within your VPC security perimeter

Question 5

Your team is working on a binary classification problem. You have trained a support vector machine (SVM) classifier with default parameters, and received an area under the Curve (AUC) of 0.87 on the validation set. You want to increase the AUC of the model. What should you do?
A. Perform hyperparameter tuning
B. Train a classifier with deep neural networks, because neural networks would always beat SVMs
C. Deploy the model and measure the real-world AUC; it’s always higher because of generalization
D. Scale predictions you get out of the model (tune a scaling factor as a hyperparameter) in order to get the highest AUC

Answer 5

A. Perform hyperparameter tuning

Question 6

You plan to deploy Cloud SQL using MySQL. You need to ensure high availability in the event of a zone failure. What should you do?
A. Create a Cloud SQL instance in one zone, and create a failover replica in another zone within the same region.
B. Create a Cloud SQL instance in one zone, and create a read replica in another zone within the same region.
C. Create a Cloud SQL instance in one zone, and configure an external read replica in a zone in a different region.
D. Create a Cloud SQL instance in a region, and configure automatic backup to a Cloud Storage bucket in the same region.

Answer 6

A. Create a Cloud SQL instance in one zone, and create a failover replica in another zone within the same region.
Reason: high availability = failover replica

Question 7

You operate an IoT pipeline built around Apache Kafka that normally receives around 5000 messages per second. You want to use Google Cloud Platform to create an alert as soon as the moving average over 1 hour drops below 4000 messages per second. What should you do?
A. Consume the stream of data in Cloud Dataflow using Kafka IO. Set a sliding time window of 1 hour every 5 minutes. Compute the average when the window closes, and send an alert if the average is less than 4000 messages.
B. Consume the stream of data in Cloud Dataflow using Kafka IO. Set a fixed time window of 1 hour. Compute the average when the window closes, and send an alert if the average is less than 4000 messages.
C. Use Kafka Connect to link your Kafka message queue to Cloud Pub/Sub. Use a Cloud Dataflow template to write your messages from Cloud Pub/Sub to Cloud Bigtable. Use Cloud Scheduler to run a script every hour that counts the number of rows created in Cloud Bigtable in the last hour. If that number falls below 4000, send an alert.
D. Use Kafka Connect to link your Kafka message queue to Cloud Pub/Sub. Use a Cloud Dataflow template to write your messages from Cloud Pub/Sub to BigQuery. Use Cloud Scheduler to run a script every five minutes that counts the number of rows created in BigQuery in the last hour. If that number falls below 4000, send an alert.

Answer 7

A. Consume the stream of data in Cloud Dataflow using Kafka IO. Set a sliding time window of 1 hour every 5 minutes. Compute the average when the window closes, and send an alert if the average is less than 4000 messages.
Reason: sliding window. C and D is not real time

Question 8

You work for an advertising company, and you’ve developed a Spark ML model to predict click-through rates at advertisement blocks. You’ve been developing everything at your on-premises data center, and now your company is migrating to Google Cloud. Your data center will be closing soon, so a rapid lift-and-shift migration is necessary. However, the data you’ve been using will be migrated to migrated to BigQuery. You periodically retrain your Spark ML models, so you need to migrate existing training pipelines to Google Cloud. What should you do?
A. Use Cloud ML Engine for training existing Spark ML models
B. Rewrite your models on TensorFlow, and start using Cloud ML Engine
C. Use Cloud Dataproc for training existing Spark ML models, but start reading data directly from BigQuery
D. Spin up a Spark cluster on Compute Engine, and train Spark ML models on the data exported from BigQuery

Answer 8

C. Use Cloud Dataproc for training existing Spark ML models, but start reading data directly from BigQuery
Reason: lift-and-shift so Dataproc.

Question 9

You want to archive data in Cloud Storage. Because some data is very sensitive, you want to use the “Trust No One” (TNO) approach to encrypt your data to prevent the cloud provider staff from decrypting your data. What should you do?
A. Use gcloud kms keys create to create a symmetric key. Then use gcloud kms encrypt to encrypt each archival file with the key and unique additional authenticated data (AAD). Use gsutil cp to upload each encrypted file to the Cloud Storage bucket, and keep the AAD outside of Google Cloud.
B. Use gcloud kms keys create to create a symmetric key. Then use gcloud kms encrypt to encrypt each archival file with the key. Use gsutil cp to upload each encrypted file to the Cloud Storage bucket. Manually destroy the key previously used for encryption, and rotate the key once.
C. Specify customer-supplied encryption key (CSEK) in the .boto configuration file. Use gsutil cp to upload each archival file to the Cloud Storage bucket. Save the CSEK in Cloud Memorystore as permanent storage of the secret.
D. Specify customer-supplied encryption key (CSEK) in the .boto configuration file. Use gsutil cp to upload each archival file to the Cloud Storage bucket. Save the CSEK in a different project that only the security team can access.

Answer 9

A. Use gcloud kms keys create to create a symmetric key. Then use gcloud kms encrypt to encrypt each archival file with the key and unique additional authenticated data (AAD). Use gsutil cp to upload each encrypted file to the Cloud Storage bucket, and keep the AAD outside of Google Cloud.
Reason: AAD is outside of Google Cloud so cloud provider staff cannot decrypt

Question 10

You have data pipelines running on BigQuery, Cloud Dataflow, and Cloud Dataproc. You need to perform health checks and monitor their behavior, and then notify the team managing the pipelines if they fail. You also need to be able to work across multiple projects. Your preference is to use managed products of features of the platform. What should you do?
A. Export the information to Cloud Stackdriver, and set up an Alerting policy
B. Run a Virtual Machine in Compute Engine with Airflow, and export the information to Stackdriver
C. Export the logs to BigQuery, and set up App Engine to read that information and send emails if you find a failure in the logs
D. Develop an App Engine application to consume logs using GCP API calls, and send emails if you find a failure in the logs

Answer 10

A. Export the information to Cloud Stackdriver, and set up an Alerting policy

Question 11

Your company is selecting a system to centralize data ingestion and delivery. You are considering messaging and data integration systems to address the requirements. The key requirements are:
✑ The ability to seek to a particular offset in a topic, possibly back to the start of all data ever captured
✑ Support for publish/subscribe semantics on hundreds of topics
✑ Retain per-key ordering
Which system should you choose?
A. Apache Kafka
B. Cloud Storage
C. Cloud Pub/Sub
D. Firebase Cloud Messaging

Answer 11

A. Apache Kafka
Reason: pub/sub max retention 31 days only

Question 12

You need to choose a database for a new project that has the following requirements:
✑ Fully managed
✑ Able to automatically scale up
✑ Transactionally consistent
✑ Able to scale up to 6 TB
✑ Able to be queried using SQL
Which database do you choose?
A. Cloud SQL
B. Cloud Bigtable
C. Cloud Spanner
D. Cloud Datastore

Answer 12

A. Cloud SQL

Reason: 6TB is ok, max 30TB for Cloud SQL

Question 13

You need to choose a database to store time series CPU and memory usage for millions of computers. You need to store this data in one-second interval samples. Analysts will be performing real-time, ad hoc analytics against the database. You want to avoid being charged for every query executed and ensure that the schema design will allow for future growth of the dataset. Which database and data model should you choose?
A. Create a table in BigQuery, and append the new samples for CPU and memory to the table
B. Create a wide table in BigQuery, create a column for the sample value at each second, and update the row with the interval for each second
C. Create a narrow table in Cloud Bigtable with a row key that combines the Computer Engine computer identifier with the sample time at each second
D. Create a wide table in Cloud Bigtable with a row key that combines the computer identifier with the sample time at each minute, and combine the values for each second as column data.

Answer 13

C. Create a narrow table in Cloud Bigtable with a row key that combines the Computer Engine computer identifier with the sample time at each second
Reason: Each row stores only 1 data point for 1 computer at a given time. Faster to write and read.

Question 14

Why do you need to split a machine learning dataset into training data and test data?
A. So you can try two different sets of features
B. To make sure your model is generalized for more than just the training data
C. To allow you to create unit tests in your code
D. So you can use one dataset for a wide model and one for a deep model

Answer 14

B. To make sure your model is generalized for more than just the training data
Reason: Other options are totally wrong

Question 15

If you want to create a machine learning model that predicts the price of a particular stock based on its recent price history, what type of estimator should you use?
A. Unsupervised learning
B. Regressor
C. Classifier
D. Clustering estimator

Answer 15

B. Regressor
Reason: Continuous value

Question 16

What Dataflow concept determines when a Window’s contents should be output based on certain criteria being met?
A. Sessions
B. OutputCriteria
C. Windows
D. Triggers

Answer 16

D. Triggers
Triggers control when the elements for a specific key and window are output. As elements arrive, they are put into one or more windows by a Window transform and its associated WindowFn, and then passed to the associated Trigger to determine if the Windows contents should be output.

Question 17

Which of the following is NOT one of the three main types of triggers that Dataflow supports?
A. Trigger based on element size in bytes
B. Trigger that is a combination of other triggers
C. Trigger based on element count
D. Trigger based on time

Answer 17

A. Trigger based on element size in bytes
Reason: There are three major kinds of triggers that Dataflow supports: 1. Time-based triggers 2. Data-driven triggers. You can set a trigger to emit results from a window when that window has received a certain number of data elements. 3. Composite triggers. These triggers combine multiple time-based or data-driven triggers in some logical way
Reference: https://cloud.google.com/dataflow/model/triggers

Question 18

You are planning to use Google’s Dataflow SDK to analyze customer data such as displayed below. Your project requirement is to extract only the customer name from the data source and then write to an output PCollection.

Tom,555 X street -

Tim,553 Y street -

Sam, 111 Z street -
Which operation is best suited for the above data processing requirement?
A. ParDo
B. Sink API
C. Source API
D. Data extraction

Answer 18

A. ParDo

Question 19

By default, which of the following windowing behavior does Dataflow apply to unbounded data sets?
A. Windows at every 100 MB of data
B. Single, Global Window
C. Windows at every 1 minute
D. Windows at every 10 minutes

Answer 19

B. Single, Global Window

Question 20

Which of these rules apply when you add preemptible workers to a Dataproc cluster (select 2 answers)?
A. Preemptible workers cannot use persistent disk.
B. Preemptible workers cannot store data.
C. If a preemptible worker is reclaimed, then a replacement worker must be added manually.
D. A Dataproc cluster cannot have only preemptible workers.

Answer 20

B. Preemptible workers cannot store data.
D. A Dataproc cluster cannot have only preemptible workers.
ReasonZ: The following rules will apply when you use preemptible workers with a Cloud Dataproc cluster:
. Processing onlySince preemptibles can be reclaimed at any time, preemptible workers do not store data. Preemptibles added to a Cloud Dataproc cluster only function as processing nodes.
. No preemptible-only clustersTo ensure clusters do not lose all workers, Cloud Dataproc cannot create preemptible-only clusters.
. Persistent disk sizeAs a default, all preemptible workers are created with the smaller of 100GB or the primary worker boot disk size. This disk space is used for local caching of data and is not available through HDFS.
The managed group automatically re-adds workers lost due to reclamation as capacity permits.
Reference: https://cloud.google.com/dataproc/docs/concepts/preemptible-vms

Question 21

Scaling a Cloud Dataproc cluster typically involves ____.
A. increasing or decreasing the number of worker nodes
B. increasing or decreasing the number of master nodes
C. moving memory to run more applications on a single node
D. deleting applications from unused nodes periodically

Answer 21

A. increasing or decreasing the number of worker nodes
Reason: After creating a Cloud Dataproc cluster, you can scale the cluster by increasing or decreasing the number of worker nodes in the cluster at any time, even when jobs are running on the cluster. Cloud Dataproc clusters are typically scaled to:
1) increase the number of workers to make a job run faster
2) decrease the number of workers to save money
3) increase the number of nodes to expand available Hadoop Distributed Filesystem (HDFS) storage
Reference: https://cloud.google.com/dataproc/docs/concepts/scaling-clusters

Question 22

The YARN ResourceManager and the HDFS NameNode interfaces are available on a Cloud Dataproc cluster ____.
A. application node
B. conditional node
C. master node
D. worker node

Answer 22

C. master node
Reason: The YARN ResourceManager and the HDFS NameNode interfaces are available on a Cloud Dataproc cluster master node. The cluster master-host-name is the name of your Cloud Dataproc cluster followed by an -m suffixfor example, if your cluster is named “my-cluster”, the master-host-name would be “my-cluster-m”.

Question 23

Which of these is NOT a way to customize the software on Dataproc cluster instances?
A. Set initialization actions
B. Modify configuration files using cluster properties
C. Configure the cluster using Cloud Deployment Manager
D. Log into the master node and make changes from there

Answer 23

C. Configure the cluster using Cloud Deployment Manager
You can access the master node of the cluster by clicking the SSH button next to it in the Cloud Console.
You can easily use the –properties option of the dataproc command in the Google Cloud SDK to modify many common configuration files when creating a cluster.
When creating a Cloud Dataproc cluster, you can specify initialization actions in executables and/or scripts that Cloud Dataproc will run on all nodes in your Cloud
Dataproc cluster immediately after the cluster is set up.

Question 24

In order to securely transfer web traffic data from your computer’s web browser to the Cloud Dataproc cluster you should use a(n) _____.
A. VPN connection
B. Special browser
C. SSH tunnel
D. FTP connection

Answer 24

C. SSH tunnel
To connect to the web interfaces, it is recommended to use an SSH tunnel to create a secure connection to the master node.

Question 25

Cloud Bigtable is Google’s ______ Big Data database service.
A. Relational
B. mySQL
C. NoSQL
D. SQL Server

Answer 25

C. NoSQL

Question 26

Cloud Bigtable is a recommended option for storing very large amounts of ____________________________?
A. multi-keyed data with very high latency
B. multi-keyed data with very low latency
C. single-keyed data with very low latency
D. single-keyed data with very high latency

Answer 26

C. single-keyed data with very low latency
Cloud Bigtable is a sparsely populated table that can scale to billions of rows and thousands of columns, allowing you to store terabytes or even petabytes of data.
A single value in each row is indexed; this value is known as the row key. Cloud Bigtable is ideal for storing very large amounts of single-keyed data with very low latency. It supports high read and write throughput at low latency, and it is an ideal data source for MapReduce operations.

Question 27

Google Cloud Bigtable indexes a single value in each row. This value is called the _______.
A. primary key
B. unique key
C. row key
D. master key

Answer 27

C. row key

Question 28

What is the HBase Shell for Cloud Bigtable?
A. The HBase shell is a GUI based interface that performs administrative tasks, such as creating and deleting tables.
B. The HBase shell is a command-line tool that performs administrative tasks, such as creating and deleting tables.
C. The HBase shell is a hypervisor based shell that performs administrative tasks, such as creating and deleting new virtualized instances.
D. The HBase shell is a command-line tool that performs only user account management functions to grant access to Cloud Bigtable instances.

Answer 28

B. The HBase shell is a command-line tool that performs administrative tasks, such as creating and deleting tables.
The HBase shell is a command-line tool that performs administrative tasks, such as creating and deleting tables. The Cloud Bigtable HBase client for Java makes it possible to use the HBase shell to connect to Cloud Bigtable.

Question 29

Does Dataflow process batch data pipelines or streaming data pipelines?
A. Only Batch Data Pipelines
B. Both Batch and Streaming Data Pipelines
C. Only Streaming Data Pipelines
D. None of the above

Answer 29

B. Both Batch and Streaming Data Pipelines

Question 30

What are all of the BigQuery operations that Google charges for?
A. Storage, queries, and streaming inserts
B. Storage, queries, and loading data from a file
C. Storage, queries, and exporting data
D. Queries and streaming inserts

Answer 30

A. Storage, queries, and streaming inserts
Google charges for storage, queries, and streaming inserts. Loading data from a file and exporting data are free operations.

Question 31

When using Cloud Dataproc clusters, you can access the YARN web interface by configuring a browser to connect through a ____ proxy.
A. HTTPS
B. VPN
C. SOCKS
D. HTTP

Answer 31

C. SOCKS
When using Cloud Dataproc clusters, configure your browser to use the SOCKS proxy. The SOCKS proxy routes data intended for the Cloud Dataproc cluster through an SSH tunnel.

Question 32

What are two methods that can be used to denormalize tables in BigQuery?
A. 1) Split table into multiple tables; 2) Use a partitioned table
B. 1) Join tables into one table; 2) Use nested repeated fields
C. 1) Use a partitioned table; 2) Join tables into one table
D. 1) Use nested repeated fields; 2) Use a partitioned table

Answer 32

B. 1) Join tables into one table; 2) Use nested repeated fields

Question 33

Dataproc clusters contain many configuration files. To update these files, you will need to use the –properties option. The format for the option is: file_prefix:property=_____.
A. details
B. value
C. null
D. id

Answer 33

B. value
To make updating files and properties easy, the –properties command uses a special format to specify the configuration file and the property and value within the file that should be updated. The formatting is as follows: file_prefix:property=value.

Question 34

Which role must be assigned to a service account used by the virtual machines in a Dataproc cluster so they can execute jobs?
A. Dataproc Worker
B. Dataproc Viewer
C. Dataproc Runner
D. Dataproc Editor

Answer 34

A. Dataproc Worker
Service accounts used with Cloud Dataproc must have Dataproc/Dataproc Worker role (or have all the permissions granted by Dataproc Worker role).

Question 35

Which of the following is not true about Dataflow pipelines?
A. Pipelines are a set of operations
B. Pipelines represent a data processing job
C. Pipelines represent a directed graph of steps
D. Pipelines can share data between instances

Answer 35

D. Pipelines can share data between instances
The data and transforms in a pipeline are unique to, and owned by, that pipeline. While your program can create multiple pipelines, pipelines cannot share data or transforms

Question 36

Which of the following is NOT true about Dataflow pipelines?
A. Dataflow pipelines are tied to Dataflow, and cannot be run on any other runner
B. Dataflow pipelines can consume data from other Google Cloud services
C. Dataflow pipelines can be programmed in Java
D. Dataflow pipelines use a unified programming model, so can work both with streaming and batch data sources

Answer 36

A. Dataflow pipelines are tied to Dataflow, and cannot be run on any other runner

Dataflow pipelines can also run on alternate runtimes like Spark and Flink, as they are built using the Apache Beam SDKs

Question 37

Which Cloud Dataflow / Beam feature should you use to aggregate data in an unbounded data source every hour based on the time when the data entered the pipeline?
A. An hourly watermark
B. An event time trigger
C. The with Allowed Lateness method
D. A processing time trigger

Answer 37

D. A processing time trigger
Reason: “when the data entered the pipeline”
When collecting and grouping data into windows, Beam uses triggers to determine when to emit the aggregated results of each window.
Processing time triggers. These triggers operate on the processing time the time when the data element is processed at any given stage in the pipeline.
Event time triggers. These triggers operate on the event time, as indicated by the timestamp on each data element. Beams default trigger is event time-based.

Question 38

You have a job that you want to cancel. It is a streaming pipeline, and you want to ensure that any data that is in-flight is processed and written to the output.
Which of the following commands can you use on the Dataflow monitoring console to stop the pipeline job?
A. Cancel
B. Drain
C. Stop
D. Finish

Answer 38

B. Drain
Using the Drain option to stop your job tells the Dataflow service to finish your job in its current state. Your job will immediately stop ingesting new data from input sources, but the Dataflow service will preserve any existing resources (such as worker instances) to finish processing and writing any buffered data in your pipeline.

Question 39

You want to use a BigQuery table as a data sink. In which writing mode(s) can you use BigQuery as a sink?
A. Both batch and streaming
B. BigQuery cannot be used as a sink
C. Only batch
D. Only streaming

Answer 39

A. Both batch and streaming

Question 40

Which of the following are feature engineering techniques? (Select 2 answers)
A. Hidden feature layers
B. Feature prioritization
C. Crossed feature columns
D. Bucketization of a continuous feature

Answer 40

C. Crossed feature columns
D. Bucketization of a continuous feature

Selecting and crafting the right set of feature columns is key to learning an effective model.
Bucketization is a process of dividing the entire range of a continuous feature into a set of consecutive bins/buckets, and then converting the original numerical feature into a bucket ID (as a categorical feature) depending on which bucket that value falls into.
Using each base feature column separately may not be enough to explain the data. To learn the differences between different feature combinations, we can add crossed feature columns to the model.

Question 41

Which of the following are examples of hyperparameters? (Select 2 answers.)
A. Number of hidden layers
B. Number of nodes in each hidden layer
C. Biases
D. Weights

Answer 41

A. Number of hidden layers
B. Number of nodes in each hidden layer

If model parameters are variables that get adjusted by training with existing data, your hyperparameters are the variables about the training process itself. For example, part of setting up a deep neural network is deciding how many “hidden” layers of nodes to use between the input layer and the output layer, as well as how many nodes each layer should use. These variables are not directly related to the training data at all. They are configuration variables. Another difference is that parameters change during a training job, while the hyperparameters are usually constant during a job.
Weights and biases are variables that get adjusted during the training process, so they are not hyperparameters

Question 42

How can you get a neural network to learn about relationships between categories in a categorical feature?
A. Create a multi-hot column
B. Create a one-hot column
C. Create a hash bucket
D. Create an embedding column

Answer 42

D. Create an embedding column

There are two problems with one-hot encoding. First, it has high dimensionality, meaning that instead of having just one value, like a continuous feature, it has many values, or dimensions. This makes computation more time-consuming, especially if a feature has a very large number of categories. The second problem is that it doesnt encode any relationships between the categories. They are completely independent from each other, so the network has no way of knowing which ones are similar to each other.
Both of these problems can be solved by representing a categorical feature with an embedding column. The idea is that each category has a smaller vector with, lets say, 5 values in it. But unlike a one-hot vector, the values are not usually 0. The values are weights, similar to the weights that are used for basic features in a neural network. The difference is that each category has a set of weights (5 of them in this case).
You can think of each value in the embedding vector as a feature of the category. So, if two categories are very similar to each other, then their embedding vectors should be very similar too.

Question 43

What are two of the benefits of using denormalized data structures in BigQuery?
A. Reduces the amount of data processed, reduces the amount of storage required
B. Increases query speed, makes queries simpler
C. Reduces the amount of storage required, increases query speed
D. Reduces the amount of data processed, increases query speed

Answer 43

B. Increases query speed, makes queries simpler

Denormalization increases query speed for tables with billions of rows because BigQuery’s performance degrades when doing JOINs on large tables, but with a denormalized data structure, you don’t have to use JOINs, since all of the data has been combined into one table. Denormalization also makes queries simpler because you do not have to use JOIN clauses.
Denormalization increases the amount of data processed and the amount of storage required because it creates redundant data.

Question 44

What are two of the characteristics of using online prediction rather than batch prediction?
A. It is optimized to handle a high volume of data instances in a job and to run more complex models.
B. Predictions are returned in the response message.
C. Predictions are written to output files in a Cloud Storage location that you specify.
D. It is optimized to minimize the latency of serving predictions.

Answer 44

B. Predictions are returned in the response message.
D. It is optimized to minimize the latency of serving predictions.

Online prediction -
.Optimized to minimize the latency of serving predictions.
.Predictions returned in the response message.

Batch prediction -
.Optimized to handle a high volume of instances in a job and to run more complex models.
.Predictions written to output files in a Cloud Storage location that you specify.

Question 45

To run a TensorFlow training job on your own computer using Cloud Machine Learning Engine, what would your command start with?
A. gcloud ml-engine local train
B. gcloud ml-engine jobs submit training
C. gcloud ml-engine jobs submit training local
D. You can’t run a TensorFlow program on your own computer using Cloud ML Engine .

Answer 45

A. gcloud ml-engine local train

gcloud ml-engine local train - run a Cloud ML Engine training job locally
This command runs the specified module in an environment similar to that of a live Cloud ML Engine Training Job.
This is especially useful in the case of testing distributed models, as it allows you to validate that you are properly interacting with the Cloud ML Engine cluster configuration.

Question 46

Which TensorFlow function can you use to configure a categorical column if you don’t know all of the possible values for that column?
A. categorical_column_with_vocabulary_list
B. categorical_column_with_hash_bucket
C. categorical_column_with_unknown_values
D. sparse_column_with_keys

Answer 46

B. categorical_column_with_hash_bucket

If you know the set of all possible feature values of a column and there are only a few of them, you can use categorical_column_with_vocabulary_list. Each key in the list will get assigned an auto-incremental ID starting from 0.
What if we don’t know the set of possible values in advance? Not a problem. We can use categorical_column_with_hash_bucket instead. What will happen is that each possible value in the feature column occupation will be hashed to an integer ID as we encounter them in training.

Question 47

Which software libraries are supported by Cloud Machine Learning Engine?
A. Theano and TensorFlow
B. Theano and Torch
C. TensorFlow
D. TensorFlow and Torch

Answer 47

C. TensorFlow
it supports Tensflow, scikit=learn & XGBoost

Question 48

The CUSTOM tier for Cloud Machine Learning Engine allows you to specify the number of which types of cluster nodes?
A. Workers
B. Masters, workers, and parameter servers
C. Workers and parameter servers
D. Parameter servers

Answer 48

C. Workers and parameter servers

The CUSTOM tier is not a set tier, but rather enables you to use your own cluster specification. When you use this tier, set values to configure your processing cluster according to these guidelines:
You must set TrainingInput.masterType to specify the type of machine to use for your master node.
You may set TrainingInput.workerCount to specify the number of workers to use.
You may set TrainingInput.parameterServerCount to specify the number of parameter servers to use.
You can specify the type of machine for the master node, but you can’t specify more than one master node.

Question 49

All Google Cloud Bigtable client requests go through a front-end server ______ they are sent to a Cloud Bigtable node.
A. before
B. after
C. only if
D. once

Answer 49

A. before
In a Cloud Bigtable architecture all client requests go through a front-end server before they are sent to a Cloud Bigtable node.
The nodes are organized into a Cloud Bigtable cluster, which belongs to a Cloud Bigtable instance, which is a container for the cluster. Each node in the cluster handles a subset of the requests to the cluster.
When additional nodes are added to a cluster, you can increase the number of simultaneous requests that the cluster can handle, as well as the maximum throughput for the entire cluster.

Question 50

When you store data in Cloud Bigtable, what is the recommended minimum amount of stored data?
A. 500 TB
B. 1 GB
C. 1 TB
D. 500 GB

Answer 50

C. 1 TB
1TB minimum, and 2TB per node