A gaming organization is developing a new game and would like to offer real-time competition to their users. The data architecture has the following characteristics:
✑ The game application is writing events directly to Amazon DynamoDB from the user's mobile device.
✑ Users from the website can access their statistics directly from DynamoDB.
✑ The game servers are accessing DynamoDB to update the user's information.
✑ The data science team extracts data from DynamoDB for various applications.
The engineering team has already agreed to the IAM roles and policies to use for the data science team and the application.
Which actions will provide the MOST security, while maintaining the necessary access to the website and game application? (Choose two.)

A. Use Amazon Cognito user pool to authenticate to both the website and the game application.

B. Use IAM identity federation to authenticate to both the website and the game application.

C. Create an IAM policy with PUT permission for both the website and the game application.

D. Create an IAM policy with fine-grained permission for both the website and the game application.

E. Create an IAM policy with PUT permission for the game application and an IAM policy with GET permission for the website.

A company hosts a portfolio of e-commerce websites across the Oregon, N. Virginia, Ireland, and Sydney
AWS regions. Each site keeps log files that capture user behavior. The company has built an application that generates batches of product recommendations with collaborative filtering in Oregon. Oregon was selected because the flagship site is hosted there and provides the largest collection of data to train machine learning models against. The other regions do NOT have enough historic data to train accurate machine learning models.
Which set of data processing steps improves recommendations for each region?

A. Use the e-commerce application in Oregon to write replica log files in each other region.

B. Use Amazon S3 bucket replication to consolidate log entries and build a single model in Oregon.

C. Use Kinesis as a buffer for web logs and replicate logs to the Kinesis stream of a neighboring region.

D. Use the CloudWatch Logs agent to consolidate logs into a single CloudWatch Logs group.

A customer is collecting clickstream data using Amazon Kinesis and is grouping the events by IP address into
5-minute chunks stored in Amazon S3.
Many analysts in the company use Hive on Amazon EMR to analyze this data. Their queries always reference a single IP address. Data must be optimized for querying based on IP address using Hive running on Amazon
EMR.
What is the most efficient method to query the data with Hive?

A. Store an index of the files by IP address in the Amazon DynamoDB metadata store for EMRFS.

B. Store the Amazon S3 objects with the following naming scheme: bucket_name/source=ip_address/ year=yy/month=mm/day=dd/hour=hh/filename.

C. Store the data in an HBase table with the IP address as the row key.

D. Store the events for an IP address as a single file in Amazon S3 and add metadata with keys: Hive_Partitioned_IPAddress.

A real-time bidding company is rebuilding their monolithic application and is focusing on serving real-time data. A large number of reads and writes are generated from thousands of concurrent users who follow items and bid on the company's sale offers.
The company is experiencing high latency during special event spikes, with millions of concurrent users.
The company needs to analyze and aggregate a part of the data in near real time to feed an internal dashboard.
What is the BEST approach for serving and analyzing data, considering the constraint of the row latency on the highly demanded data?

A. Use Amazon Aurora with Multi Availability Zone and read replicas. Use Amazon ElastiCache in front of the read replicas to serve read-only content quickly. Use the same database as datasource for the dashboard.

B. Use Amazon DynamoDB to store real-time data with Amazon DynamoDB. Accelerator to serve content quickly. use Amazon DynamoDB Streams to replay all changes to the table, process and stream to Amazon Elasti search Service with AWS Lambda.

C. Use Amazon RDS with Multi Availability Zone. Provisioned IOPS EBS volume for storage. Enable up to five read replicas to serve read-only content quickly. Use Amazon EMR with Sqoop to import Amazon RDS data into HDFS for analysis.

D. Use Amazon Redshift with a DC2 node type and a multi-mode cluster. Create an Amazon EC2 instance with pgpoo1 installed. Create an Amazon ElastiCache cluster and route read requests through pgpoo1, and use Amazon Redshift for analysis. D

An organization currently runs a large Hadoop environment in their data center and is in the process of creating an alternative Hadoop environment on AWS, using Amazon EMR.
They generate around 20 TB of data on a monthly basis. Also on a monthly basis, files need to be grouped and copied to Amazon S3 to be used for the Amazon
EMR environment. They have multiple S3 buckets across AWS accounts to which data needs to be copied. There is a 10G AWS Direct Connect setup between their data center and AWS, and the network team has agreed to allocate 50% of AWS Direct Connect bandwidth to data transfer. The data transfer cannot take more than two days.
What would be the MOST efficient approach to transfer data to AWS on a monthly basis?

A. Use an offline copy method, such as an AWS Snowball device, to copy and transfer data to Amazon S3.

B. Configure a multipart upload for Amazon S3 on AWS Java SDK to transfer data over AWS Direct Connect.

C. Use Amazon S3 transfer acceleration capability to transfer data to Amazon S3 over AWS Direct Connect.

D. Setup S3DistCop tool on the on-premises Hadoop environment to transfer data to Amazon S3 over AWS Direct Connect.

An organization is soliciting public feedback through a web portal that has been deployed to track the number of requests and other important data. As part of reporting and visualization, AmazonQuickSight connects to an Amazon RDS database to virtualize data. Management wants to understand some important metrics about feedback and how the feedback has changed over the last four weeks in a visual representation.
What would be the MOST effective way to represent multiple iterations of an analysis in Amazon QuickSight that would show how the data has changed over the last four weeks?

A. Use the analysis option for data captured in each week and view the data by a date range.

B. Use a pivot table as a visual option to display measured values and weekly aggregate data as a row dimension.

C. Use a dashboard option to create an analysis of the data for each week and apply filters to visualize the data change.

D. Use a story option to preserve multiple iterations of an analysis and play the iterations sequentially.

A telecommunications company needs to predict customer churn (i.e., customers who decide to switch to a competitor). The company has historic records of each customer, including monthly consumption patterns, calls to customer service, and whether the customer ultimately quit the service. All of this data is stored in
Amazon S3. The company needs to know which customers are likely going to churn soon so that they can win back their loyalty.
What is the optimal approach to meet these requirements?

A. Use the Amazon Machine Learning service to build the binary classification model based on the dataset stored in Amazon S3. The model will be used regularly to predict churn attribute for existing customers.

B. Use AWS QuickSight to connect it to data stored in Amazon S3 to obtain the necessary business insight. Plot the churn trend graph to extrapolate churn likelihood for existing customers.

C. Use EMR to run the Hive queries to build a profile of a churning customer. Apply a profile to existing customers to determine the likelihood of churn.

D. Use a Redshift cluster to COPY the data from Amazon S3. Create a User Defined Function in Redshift that computes the likelihood of churn.

A company that manufactures and sells smart air conditioning units also offers add-on services so that customers can see real-time dashboards in a mobile application or a web browser. Each unit sends its sensor information in JSON format every two seconds for processing and analysis. The company also needs to consume this data to predict possible equipment problems before they occur. A few thousand pre-purchased units will be delivered in the next couple of months. The company expects high market growth in the next year and needs to handle a massive amount of data and scale without interruption.
Which ingestion solution should the company use?

A. Write sensor data records to Amazon Kinesis Streams. Process the data using KCL applications for the end-consumer dashboard and anomaly detection workflows.

B. Batch sensor data to Amazon Simple Storage Service (S3) every 15 minutes. Flow the data downstream to the end-consumer dashboard and to the anomaly detection application.

C. Write sensor data records to Amazon Kinesis Firehose with Amazon Simple Storage Service (S3) as the destination. Consume the data with a KCL application for the end-consumer dashboard and anomaly detection.

D. Write sensor data records to Amazon Relational Database Service (RDS). Build both the end-consumer dashboard and anomaly detection application on top of Amazon RDS.

An Amazon EMR cluster using EMRFS has access to petabytes of data on Amazon S3, originating from multiple unique data sources. The customer needs to query common fields across some of the data sets to be able to perform interactive joins and then display results quickly.
Which technology is most appropriate to enable this capability?

A. Presto

B. MicroStrategy

C. Pig

D. R Studio

An administrator needs to design the event log storage architecture for events from mobile devices. The event data will be processed by an Amazon EMR cluster daily for aggregated reporting and analytics before being archived.
How should the administrator recommend storing the log data?

A. Create an Amazon S3 bucket and write log data into folders by device. Execute the EMR job on the device folders.

B. Create an Amazon DynamoDB table partitioned on the device and sorted on date, write log data to table. Execute the EMR job on the Amazon DynamoDB table.

C. Create an Amazon S3 bucket and write data into folders by day. Execute the EMR job on the daily folder.

D. Create an Amazon DynamoDB table partitioned on EventID, write log data to table. Execute the EMR job on the table.

A new algorithm has been written in Python to identify SPAM e-mails. The algorithm analyzes the free text contained within a sample set of 1 million e-mails stored on Amazon S3. The algorithm must be scaled across a production dataset of 5 PB, which also resides in Amazon S3 storage.
Which AWS service strategy is best for this use case?

A. Copy the data into Amazon ElastiCache to perform text analysis on the in-memory data and export the results of the model into Amazon Machine Learning.

B. Use Amazon EMR to parallelize the text analysis tasks across the cluster using a streaming program step.

C. Use Amazon Elasticsearch Service to store the text and then use the Python Elasticsearch Client to run analysis against the text index.

D. Initiate a Python job from AWS Data Pipeline to run directly against the Amazon S3 text files.

A large oil and gas company needs to provide near real-time alerts when peak thresholds are exceeded in its pipeline system. The company has developed a system to capture pipeline metrics such as flow rate, pressure, and temperature using millions of sensors. The sensors deliver to AWS IoT.
What is a cost-effective way to provide near real-time alerts on the pipeline metrics?

A. Create an AWS IoT rule to generate an Amazon SNS notification.

B. Store the data points in an Amazon DynamoDB table and poll if for peak metrics data from an Amazon EC2 application.

C. Create an Amazon Machine Learning model and invoke it with AWS Lambda.

D. Use Amazon Kinesis Streams and a KCL-based application deployed on AWS Elastic Beanstalk.

A game company needs to properly scale its game application, which is backed by DynamoDB. Amazon
Redshift has the past two years of historical data. Game traffic varies throughout the year based on various factors such as season, movie release, and holiday season. An administrator needs to calculate how much read and write throughput should be provisioned for DynamoDB table for each week in advance.
How should the administrator accomplish this task?

A. Feed the data into Amazon Machine Learning and build a regression model.

B. Feed the data into Spark Mlib and build a random forest modest.

C. Feed the data into Apache Mahout and build a multi-classification model.

D. Feed the data into Amazon Machine Learning and build a binary classification model.

An organization is designing an application architecture. The application will have over 100 TB of data and will support transactions that arrive at rates from hundreds per second to tens of thousands per second, depending on the day of the week and time of the day. All transaction data, must be durably and reliably stored. Certain read operations must be performed with strong consistency.
Which solution meets these requirements?

A. Use Amazon DynamoDB as the data store and use strongly consistent reads when necessary.

B. Use an Amazon Relational Database Service (RDS) instance sized to meet the maximum anticipated transaction rate and with the High Availability option enabled.

C. Deploy a NoSQL data store on top of an Amazon Elastic MapReduce (EMR) cluster, and select the HDFS High Durability option.

D. Use Amazon Redshift with synchronous replication to Amazon Simple Storage Service (S3) and row-level locking for strong consistency.

A web-hosting company is building a web analytics tool to capture clickstream data from all of the websites hosted within its platform and to provide near-real-time business intelligence. This entire system is built on
AWS services. The web-hosting company is interested in using Amazon Kinesis to collect this data and perform sliding window analytics.
What is the most reliable and fault-tolerant technique to get each website to send data to Amazon Kinesis with every click?

A. After receiving a request, each web server sends it to Amazon Kinesis using the Amazon Kinesis PutRecord API. Use the sessionID as a partition key and set up a loop to retry until a success response is received.

B. After receiving a request, each web server sends it to Amazon Kinesis using the Amazon Kinesis Producer Library .addRecords method.

C. Each web server buffers the requests until the count reaches 500 and sends them to Amazon Kinesis using the Amazon Kinesis PutRecord API.

D. After receiving a request, each web server sends it to Amazon Kinesis using the Amazon Kinesis PutRecord API. Use the exponential back-off algorithm for retries until a successful response is received.

A company is building a new application in AWS. The architect needs to design a system to collect application log events. The design should be a repeatable pattern that minimizes data loss if an application instance fails, and keeps a durable copy of a log data for at least 30 days.
What is the simplest architecture that will allow the architect to analyze the logs?

A. Write them directly to a Kinesis Firehose. Configure Kinesis Firehose to load the events into an Amazon Redshift cluster for analysis.

B. Write them to a file on Amazon Simple Storage Service (S3). Write an AWS Lambda function that runs in response to the S3 event to load the events into Amazon Elasticsearch Service for analysis.

C. Write them to the local disk and configure the Amazon CloudWatch Logs agent to load the data into CloudWatch Logs and subsequently into Amazon Elasticsearch Service.

D. Write them to CloudWatch Logs and use an AWS Lambda function to load them into HDFS on an Amazon Elastic MapReduce (EMR) cluster for analysis.

How should an Administrator BEST architect a large multi-layer Long Short-Term Memory (LSTM) recurrent neural network (RNN) running with MXNET on
Amazon EC2? (Choose two.)

A. Use data parallelism to partition the workload over multiple devices and balance the workload within the GPUs.

B. Use compute-optimized EC2 instances with an attached elastic GPU.

C. Use general purpose GPU computing instances such as G3 and P3.

D. Use processing parallelism to partition the workload over multiple storage devices and balance the workload within the GPUs.

An organization uses Amazon Elastic MapReduce(EMR) to process a series of extract-transform-load (ETL) steps that run in sequence. The output of each step must be fully processed in subsequent steps but will not be retained.
Which of the following techniques will meet this requirement most efficiently?

A. Use the EMR File System (EMRFS) to store the outputs from each step as objects in Amazon Simple Storage Service (S3).

B. Use the s3n URI to store the data to be processed as objects in Amazon S3.

C. Define the ETL steps as separate AWS Data Pipeline activities.

D. Load the data to be processed into HDFS, and then write the final output to Amazon S3.

An administrator needs to manage a large catalog of items from various external sellers. The administrator needs to determine if the items should be identified as minimally dangerous, dangerous, or highly dangerous based on their textual descriptions. The administrator already has some items with the danger attribute, but receives hundreds of new item descriptions every day without such classification.
The administrator has a system that captures dangerous goods reports from customer support team of from user feedback.
What is a cost-effective architecture to solve this issue?

A. Build a set of regular expression rules that are based on the existing examples, and run them on the DynamoDB Streams as every new item description is added to the system.

B. Build a Kinesis Streams process that captures and marks the relevant items in the dangerous goods reports using a Lambda function once more than two reports have been filed.

C. Build a machine learning model to properly classify dangerous goods and run it on the DynamoDB Streams as every new item description is added to the system.

D. Build a machine learning model with binary classification for dangerous goods and run it on the DynamoDB Streams as every new item description is added to the system.

A company operates an international business served from a single AWS region. The company wants to expand into a new country. The regulator for that country requires the Data Architect to maintain a log of financial transactions in the country within 24 hours of the product transaction. The production application is latency insensitive. The new country contains another AWS region.
What is the most cost-effective way to meet this requirement?

A. Use CloudFormation to replicate the production application to the new region.

B. Use Amazon CloudFront to serve application content locally in the country; Amazon CloudFront logs will satisfy the requirement.

C. Continue to serve customers from the existing region while using Amazon Kinesis to stream transaction data to the regulator.

D. Use Amazon S3 cross-region replication to copy and persist production transaction logs to a bucket in the new countrys region.

An organization is currently using an Amazon EMR long-running cluster with the latest Amazon EMR release for analytic jobs and is storing data as external tables on Amazon S3.
The company needs to launch multiple transient EMR clusters to access the same tables concurrently, but the metadata about the Amazon S3 external tables are defined and stored on the long-running cluster.
Which solution will expose the Hive metastore with the LEAST operational effort?

A. Export Hive metastore information to Amazon DynamoDB hive-site classification to point to the Amazon DynamoDB table.

B. Export Hive metastore information to a MySQL table on Amazon RDS and configure the Amazon EMR hive-site classification to point to the Amazon RDS database.

C. Launch an Amazon EC2 instance, install and configure Apache Derby, and export the Hive metastore information to derby.

D. Create and configure an AWS Glue Data Catalog as a Hive metastore for Amazon EMR.

A data engineer is running a DWH on a 25-node Redshift cluster of a SaaS service. The data engineer needs to build a dashboard that will be used by customers. Five big customers represent 80% of usage, and there is a long tail of dozens of smaller customers. The data engineer has selected the dashboarding tool.
How should the data engineer make sure that the larger customer workloads do NOT interfere with the smaller customer workloads?

A. Apply query filters based on customer-id that can NOT be changed by the user and apply distribution keys on customer-id.

B. Place the largest customers into a single user group with a dedicated query queue and place the rest of the customers into a different query queue.

C. Push aggregations into an RDS for Aurora instance. Connect the dashboard application to Aurora rather than Redshift for faster queries.

D. Route the largest customers to a dedicated Redshift cluster. Raise the concurrency of the multi-tenant Redshift cluster to accommodate the remaining customers.

A clinical trial will rely on medical sensors to remotely assess patient health. Each physician who participates in the trial requires visual reports each morning. The reports are built from aggregations of all the sensor data taken each minute.
What is the most cost-effective solution for creating this visualization each day?

A. Use Kinesis Aggregators Library to generate reports for reviewing the patient sensor data and generate a QuickSight visualization on the new data each morning for the physician to review.

B. Use a transient EMR cluster that shuts down after use to aggregate the patient sensor data each night and generate a QuickSight visualization on the new data each morning for the physician to review.

C. Use Spark streaming on EMR to aggregate the patient sensor data in every 15 minutes and generate a QuickSight visualization on the new data each morning for the physician to review.

D. Use an EMR cluster to aggregate the patient sensor data each night and provide Zeppelin notebooks that look at the new data residing on the cluster each morning for the physician to review.

An Amazon Redshift Database is encrypted using KMS. A data engineer needs to use the AWS CLI to create a KMS encrypted snapshot of the database in another AWS region.
Which three steps should the data engineer take to accomplish this task? (Choose three.)

A. Create a new KMS key in the destination region.

B. Copy the existing KMS key to the destination region.

C. Use CreateSnapshotCopyGrant to allow Amazon Redshift to use the KMS key from the source region.

D. In the source region, enable cross-region replication and specify the name of the copy grant created.

E. In the destination region, enable cross-region replication and specify the name of the copy grant created.

F. Use CreateSnapshotCopyGrant to allow Amazon Redshift to use the KMS key created in the destination region. ADF

An organization is developing a mobile social application and needs to collect logs from all devices on which it is installed. The organization is evaluating the
Amazon Kinesis Data Streams to push logs and Amazon EMR to process data. They want to store data on HDFS using the default replication factor to replicate data among the cluster, but they are concerned about the durability of the data. Currently, they are producing 300 GB of raw data daily, with additional spikes during special events. They will need to scale out the Amazon EMR cluster to match the increase in streamed data.
Which solution prevents data loss and matches compute demand?

A. Use multiple Amazon EBS volumes on Amazon EMR to store processed data and scale out the Amazon EMR cluster as needed.

B. Use the EMR File System and Amazon S3 to store processed data and scale out the Amazon EMR cluster as needed.

C. Use Amazon DynamoDB to store processed data and scale out the Amazon EMR cluster as needed.

D. use Amazon Kinesis Data Firehose and, instead of using Amazon EMR, stream logs directly into Amazon Elasticsearch Service.

An online photo album app has a key design feature to support multiple screens (e.g, desktop, mobile phone, and tablet) with high-quality displays. Multiple versions of the image must be saved in different resolutions and layouts.
The image-processing Java program takes an average of five seconds per upload, depending on the image size and format. Each image upload captures the following image metadata: user, album, photo label, upload timestamp.
The app should support the following requirements:
✑ Hundreds of user image uploads per second
✑ Maximum image upload size of 10 MB
✑ Maximum image metadata size of 1 KB
✑ Image displayed in optimized resolution in all supported screens no later than one minute after image upload
Which strategy should be used to meet these requirements?

A. Write images and metadata to Amazon Kinesis. Use a Kinesis Client Library (KCL) application to run the image processing and save the image output to Amazon S3 and metadata to the app repository DB.

B. Write image and metadata RDS with BLOB data type. Use AWS Data Pipeline to run the image processing and save the image output to Amazon S3 and metadata to the app repository DB.

C. Upload image with metadata to Amazon S3, use Lambda function to run the image processing and save the images output to Amazon S3 and metadata to the app repository DB.

D. Write image and metadata to Amazon Kinesis. Use Amazon Elastic MapReduce (EMR) with Spark Streaming to run image processing and save the images output to Amazon S3 and metadata to app repository DB.

An Operations team continuously monitors the number of visitors to a website to identify any potential system problems. The number of website visitors varies throughout the day. The site is more popular in the middle of the day and less popular at night.
Which type of dashboard display would be the MOST useful to allow staff to quickly and correctly identify system problems?

A. A vertical stacked bar chart showing today’s website visitors and the historical average number of website visitors.

B. An overlay line chart showing today’s website visitors at one-minute intervals and also the historical average number of website visitors.

C. A single KPI metric showing the statistical variance between the current number of website visitors and the historical number of website visitors for the current time of day.

D. A scatter plot showing today’s website visitors on the X-axis and the historical average number of website visitors on the Y-axis.

An organization would like to run analytics on their Elastic Load Balancing logs stored in Amazon S3 and join this data with other tables in Amazon S3. The users are currently using a BI tool connecting with JDBC and would like to keep using this BI tool.
Which solution would result in the LEAST operational overhead?

A. Trigger a Lambda function when a new log file is added to the bucket to transform and load it into Amazon Redshift. Run the VACUUM command on the Amazon Redshift cluster every night.

B. Launch a long-running Amazon EMR cluster that continuously downloads and transforms new files from Amazon S3 into its HDFS storage. Use Presto to expose the data through JDBC.

C. Trigger a Lambda function when a new log file is added to the bucket to transform and move it to another bucket with an optimized data structure. Use Amazon Athena to query the optimized bucket.

D. Launch a transient Amazon EMR cluster every night that transforms new log files and loads them into Amazon Redshift.

An organization is setting up a data catalog and metadata management environment for their numerous data stores currently running on AWS. The data catalog will be used to determine the structure and other attributes of data in the data stores. The data stores are composed of Amazon RDS databases, Amazon
Redshift, and CSV files residing on Amazon S3. The catalog should be populated on a scheduled basis, and minimal administration is required to manage the catalog.
How can this be accomplished?

A. Set up Amazon DynamoDB as the data catalog and run a scheduled AWS Lambda function that connects to data sources to populate the DynamoDB table.

B. Use an Amazon database as the data catalog and run a scheduled AWS Lambda function that connects to data sources to populate the database.

C. Use AWS Glue Data Catalog as the data catalog and schedule crawlers that connect to data sources to populate the catalog.

D. Set up Apache Hive metastore on an Amazon EC2 instance and run a scheduled bash script that connects to data sources to populate the metastore.

A customer has an Amazon S3 bucket. Objects are uploaded simultaneously by a cluster of servers from multiple streams of data. The customer maintains a catalog of objects uploaded in Amazon S3 using an
Amazon DynamoDB table. This catalog has the following fileds: StreamName, TimeStamp, and ServerName, from which ObjectName can be obtained.
The customer needs to define the catalog to support querying for a given stream or server within a defined time range.
Which DynamoDB table scheme is most efficient to support these queries?

A. Define a Primary Key with ServerName as Partition Key and TimeStamp as Sort Key. Do NOT define a Local Secondary Index or Global Secondary Index.

B. Define a Primary Key with StreamName as Partition Key and TimeStamp followed by ServerName as Sort Key. Define a Global Secondary Index with ServerName as partition key and TimeStamp followed by StreamName.

C. Define a Primary Key with ServerName as Partition Key. Define a Local Secondary Index with StreamName as Partition Key. Define a Global Secondary Index with TimeStamp as Partition Key.

D. Define a Primary Key with ServerName as Partition Key. Define a Local Secondary Index with TimeStamp as Partition Key. Define a Global Secondary Index with StreamName as Partition Key and TimeStamp as Sort Key.

An organization's data warehouse contains sales data for reporting purposes. data governance policies prohibit staff from accessing the customers' credit card numbers.
How can these policies be adhered to and still allow a Data Scientist to group transactions that use the same credit card number?

A. Store a cryptographic hash of the credit card number.

B. Encrypt the credit card number with a symmetric encryption key, and give the key only to the authorized Data Scientist.

C. Mask the credit card numbers to only show the last four digits of the credit card number.

D. Encrypt the credit card number with an asymmetric encryption key and give the decryption key only to the authorized Data Scientist.

A customer needs to determine the optimal distribution strategy for the ORDERS fact table in its Redshift schema. The ORDERS table has foreign key relationships with multiple dimension tables in this schema.
How should the company determine the most appropriate distribution key for the ORDERS table?

A. Identify the largest and most frequently joined dimension table and ensure that it and the ORDERS table both have EVEN distribution.

B. Identify the largest dimension table and designate the key of this dimension table as the distribution key of the ORDERS table.

C. Identify the smallest dimension table and designate the key of this dimension table as the distribution key of the ORDERS table.

D. Identify the largest and the most frequently joined dimension table and designate the key of this dimension table as the distribution key of the ORDERS table.

An Amazon Kinesis stream needs to be encrypted.
Which approach should be used to accomplish this task?

A. Perform a client-side encryption of the data before it enters the Amazon Kinesis stream on the producer.

B. Use a partition key to segment the data by MD5 hash function, which makes it undecipherable while in transit.

C. Perform a client-side encryption of the data before it enters the Amazon Kinesis stream on the consumer.

D. Use a shard to segment the data, which has built-in functionality to make it indecipherable while in transit.

An administrator needs to design a distribution strategy for a star schema in a Redshift cluster. The administrator needs to determine the optimal distribution style for the tables in the Redshift schema.
In which three circumstances would choosing Key-based distribution be most appropriate? (Select three.)

A. When the administrator needs to optimize a large, slowly changing dimension table.

B. When the administrator needs to reduce cross-node traffic.

C. When the administrator needs to optimize the fact table for parity with the number of slices.

D. When the administrator needs to balance data distribution and collocation data.

E. When the administrator needs to take advantage of data locality on a local node for joins and aggregates.

A company receives data sets coming from external providers on Amazon S3. Data sets from different providers are dependent on one another. Data sets will arrive at different times and in no particular order.
A data architect needs to design a solution that enables the company to do the following:
✑ Rapidly perform cross data set analysis as soon as the data becomes available
✑ Manage dependencies between data sets that arrive at different times
Which architecture strategy offers a scalable and cost-effective solution that meets these requirements?

A. Maintain data dependency information in Amazon RDS for MySQL. Use an AWS Data Pipeline job to load an Amazon EMR Hive table based on task dependencies and event notification triggers in Amazon S3.

B. Maintain data dependency information in an Amazon DynamoDB table. Use Amazon SNS and event notifications to publish data to fleet of Amazon EC2 workers. Once the task dependencies have been resolved, process the data with Amazon EMR.

C. Maintain data dependency information in an Amazon ElastiCache Redis cluster. Use Amazon S3 event notifications to trigger an AWS Lambda function that maps the S3 object to Redis. Once the task dependencies have been resolved, process the data with Amazon EMR.

D. Maintain data dependency information in an Amazon DynamoDB table. Use Amazon S3 event notifications to trigger an AWS Lambda function that maps the S3 object to the task associated with it in DynamoDB. Once all task dependencies have been resolved, process the data with Amazon EMR.

An organization uses a custom map reduce application to build monthly reports based on many small data files in an Amazon S3 bucket. The data is submitted from various business units on a frequent but unpredictable schedule. As the dataset continues to grow, it becomes increasingly difficult to process all of the data in one day. The organization has scaled up its Amazon EMR cluster, but other optimizations could improve performance.
The organization needs to improve performance with minimal changes to existing processes and applications.
What action should the organization take?

A. Use Amazon S3 Event Notifications and AWS Lambda to create a quick search file index in DynamoDB.

B. Add Spark to the Amazon EMR cluster and utilize Resilient Distributed Datasets in-memory.

C. Use Amazon S3 Event Notifications and AWS Lambda to index each file into an Amazon Elasticsearch Service cluster.

D. Schedule a daily AWS Data Pipeline process that aggregates content into larger files using S3DistCp.

E. Have business units submit data via Amazon Kinesis Firehose to aggregate data hourly into Amazon S3.

An organization has 10,000 devices that generate 100 GB of telemetry data per day, with each record size around 10 KB. Each record has 100 fields, and one field consists of unstructured log data with a "String" data type in the English language. Some fields are required for the real-time dashboard, but all fields must be available for long-term generation.
The organization also has 10 PB of previously cleaned and structured data, partitioned by Date, in a SAN that must be migrated to AWS within one month.
Currently, the organization does not have any real-time capabilities in their solution. Because of storage limitations in the on-premises data warehouse, selective data is loaded while generating the long-term trend with ANSI SQL queries through JDBC for visualization. In addition to the one-time data loading, the organization needs a cost-effective and real-time solution.
How can these requirements be met? (Choose two.)

A. use AWS IoT to send data from devices to an Amazon SQS queue, create a set of workers in an Auto Scaling group and read records in batch from the queue to process and save the data. Fan out to an Amazon SNS queue attached with an AWS Lambda function to filter the request dataset and save it to Amazon Elasticsearch Service for real-time analytics.

B. Create a Direct Connect connection between AWS and the on-premises data center and copy the data to Amazon S3 using S3 Acceleration. Use Amazon Athena to query the data.

C. Use AWS IoT to send the data from devices to Amazon Kinesis Data Streams with the IoT rules engine. Use one Kinesis Data Firehose stream attached to a Kinesis stream to batch and stream the data partitioned by date. Use another Kinesis Firehose stream attached to the same Kinesis stream to filter out the required fields to ingest into Elasticsearch for real-time analytics.

D. Use AWS IoT to send the data from devices to Amazon Kinesis Data Streams with the IoT rules engine. Use one Kinesis Data Firehose stream attached to a Kinesis stream to stream the data into an Amazon S3 bucket partitioned by date. Attach an AWS Lambda function with the same Kinesis stream to filter out the required fields for ingestion into Amazon DynamoDB for real-time analytics.

E. use multiple AWS Snowball Edge devices to transfer data to Amazon S3, and use Amazon Athena to query the data.

A data engineer is about to perform a major upgrade to the DDL contained within an Amazon Redshift cluster to support a new data warehouse application. The upgrade scripts will include user permission updates, view and table structure changes as well as additional loading and data manipulation tasks.
The data engineer must be able to restore the database to its existing state in the event of issues.
Which action should be taken prior to performing this upgrade task?

A. Run an UNLOAD command for all data in the warehouse and save it to S3.

B. Create a manual snapshot of the Amazon Redshift cluster.

C. Make a copy of the automated snapshot on the Amazon Redshift cluster.

D. Call the waitForSnapshotAvailable command from either the AWS CLI or an AWS SDK.

There are thousands of text files on Amazon S3. The total size of the files is 1 PB. The files contain retail order information for the past 2 years. A data engineer needs to run multiple interactive queries to manipulate the data. The Data Engineer has AWS access to spin up an Amazon EMR cluster. The data engineer needs to use an application on the cluster to process this data and return the results in interactive time frame.
Which application on the cluster should the data engineer use?

A. Oozie

B. Apache Pig with Tachyon

C. Apache Hive

D. Presto

The department of transportation for a major metropolitan area has placed sensors on roads at key locations around the city. The goal is to analyze the flow of traffic and notifications from emergency services to identify potential issues and to help planners correct trouble spots.
A data engineer needs a scalable and fault-tolerant solution that allows planners to respond to issues within
30 seconds of their occurrence.
Which solution should the data engineer choose?

A. Collect the sensor data with Amazon Kinesis Firehose and store it in Amazon Redshift for analysis. Collect emergency services events with Amazon SQS and store in Amazon DynampDB for analysis.

B. Collect the sensor data with Amazon SQS and store in Amazon DynamoDB for analysis. Collect emergency services events with Amazon Kinesis Firehose and store in Amazon Redshift for analysis.

C. Collect both sensor data and emergency services events with Amazon Kinesis Streams and use DynamoDB for analysis.

D. Collect both sensor data and emergency services events with Amazon Kinesis Firehose and use Amazon Redshift for analysis.

An advertising organization uses an application to process a stream of events that are received from clients in multiple unstructured formats.
The application does the following:
✑ Transforms the events into a single structured format and streams them to Amazon Kinesis for real-time analysis.
✑ Stores the unstructured raw events from the log files on local hard drivers that are rotated and uploaded to Amazon S3.
The organization wants to extract campaign performance reporting using an existing Amazon redshift cluster.
Which solution will provide the performance data with the LEAST number of operations?

A. Install the Amazon Kinesis Data Firehose agent on the application servers and use it to stream the log files directly to Amazon Redshift.

B. Create an external table in Amazon Redshift and point it to the S3 bucket where the unstructured raw events are stored.

C. Write an AWS Lambda function that triggers every hour to load the new log files already in S3 to Amazon redshift.

D. Connect Amazon Kinesis Data Firehose to the existing Amazon Kinesis stream and use it to stream the event directly to Amazon Redshift.

An organization needs to design and deploy a large-scale data storage solution that will be highly durable and highly flexible with respect to the type and structure of data being stored. The data to be stored will be sent or generated from a variety of sources and must be persistently available for access and processing by multiple applications.
What is the most cost-effective technique to meet these requirements?

A. Use Amazon Simple Storage Service (S3) as the actual data storage system, coupled with appropriate tools for ingestion/acquisition of data and for subsequent processing and querying.

B. Deploy a long-running Amazon Elastic MapReduce (EMR) cluster with Amazon Elastic Block Store (EBS) volumes for persistent HDFS storage and appropriate Hadoop ecosystem tools for processing and querying.

C. Use Amazon Redshift with data replication to Amazon Simple Storage Service (S3) for comprehensive durable data storage, processing, and querying.

D. Launch an Amazon Relational Database Service (RDS), and use the enterprise grade and capacity of the Amazon Aurora engine for storage, processing, and querying.

An enterprise customer is migrating to Redshift and is considering using dense storage nodes in its Redshift cluster. The customer wants to migrate 50 TB of data. The customers query patterns involve performing many joins with thousands of rows.
The customer needs to know how many nodes are needed in its target Redshift cluster. The customer has a limited budget and needs to avoid performing tests unless absolutely needed.
Which approach should this customer use?

A. Start with many small nodes.

B. Start with fewer large nodes.

C. Have two separate clusters with a mix of a small and large nodes.

D. Insist on performing multiple tests to determine the optimal configuration.

An administrator is processing events in near real-time using Kinesis streams and Lambda. Lambda intermittently fails to process batches from one of the shards due to a 5-munite time limit.
What is a possible solution for this problem?

A. Add more Lambda functions to improve concurrent batch processing.

B. Reduce the batch size that Lambda is reading from the stream.

C. Ignore and skip events that are older than 5 minutes and put them to Dead Letter Queue (DLQ).

D. Configure Lambda to read from fewer shards in parallel.

A company generates a large number of files each month and needs to use AWS import/export to move these files into Amazon S3 storage. To satisfy the auditors, the company needs to keep a record of which files were imported into Amazon S3.
What is a low-cost way to create a unique log for each import job?

A. Use the same log file prefix in the import/export manifest files to create a versioned log file in Amazon S3 for all imports.

B. Use the log file prefix in the import/export manifest files to create a unique log file in Amazon S3 for each import.

C. Use the log file checksum in the import/export manifest files to create a unique log file in Amazon S3 for each import.

D. Use a script to iterate over files in Amazon S3 to generate a log after each import/export job.

An administrator is deploying Spark on Amazon EMR for two distinct use cases: machine learning algorithms and ad-hoc querying. All data will be stored in Amazon S3. Two separate clusters for each use case will be deployed. The data volumes on Amazon S3 are less than 10 GB.
How should the administrator align instance types with the clusters purpose?

A. Machine Learning on C instance types and ad-hoc queries on R instance types

B. Machine Learning on R instance types and ad-hoc queries on G2 instance types

C. Machine Learning on T instance types and ad-hoc queries on M instance types

D. Machine Learning on D instance types and ad-hoc queries on I instance types

A city has been collecting data on its public bicycle share program for the past three years. The 5PB dataset currently resides on Amazon S3. The data contains the following datapoints:
✑ Bicycle origination points
✑ Bicycle destination points
✑ Mileage between the points
✑ Number of bicycle slots available at the station (which is variable based on the station location)
✑ Number of slots available and taken at a given time
The program has received additional funds to increase the number of bicycle stations available. All data is regularly archived to Amazon Glacier.
The new bicycle stations must be located to provide the most riders access to bicycles.
How should this task be performed?

A. Move the data from Amazon S3 into Amazon EBS-backed volumes and use an EC-2 based Hadoop cluster with spot instances to run a Spark job that performs a stochastic gradient descent optimization.

B. Use the Amazon Redshift COPY command to move the data from Amazon S3 into Redshift and perform a SQL query that outputs the most popular bicycle stations.

C. Persist the data on Amazon S3 and use a transient EMR cluster with spot instances to run a Spark streaming job that will move the data into Amazon Kinesis.

D. Keep the data on Amazon S3 and use an Amazon EMR-based Hadoop cluster with spot instances to run a Spark job that performs a stochastic gradient descent optimization over EMRFS.

An organization needs a data store to handle the following data types and access patterns:
✑ Faceting
✑ Search
✑ Flexible schema (JSON) and fixed schema
✑ Noise word elimination
Which data store should the organization choose?

A. Amazon Relational Database Service (RDS)

B. Amazon Redshift

C. Amazon DynamoDB

D. Amazon Elasticsearch Service

A media advertising company handles a large number of real-time messages sourced from over 200 websites in real time. Processing latency must be kept low. Based on calculations, a 60-shard Amazon Kinesis stream is more than sufficient to handle the maximum data throughput, even with traffic spikes. The company also uses an Amazon Kinesis Client Library (KCL) application running on Amazon Elastic Compute Cloud (EC2) managed by an Auto Scaling group. Amazon CloudWatch indicates an average of 25% CPU and a modest level of network traffic across all running servers.
The company reports a 150% to 200% increase in latency of processing messages from Amazon Kinesis during peak times. There are NO reports of delay from the sites publishing to Amazon Kinesis.
What is the appropriate solution to address the latency?

A. Increase the number of shards in the Amazon Kinesis stream to 80 for greater concurrency.

B. Increase the size of the Amazon EC2 instances to increase network throughput.

C. Increase the minimum number of instances in the Auto Scaling group.

D. Increase Amazon DynamoDB throughput on the checkpoint table.

A large grocery distributor receives daily depletion reports from the field in the form of gzip archives od CSV files uploaded to Amazon S3. The files range from 500MB to 5GB. These files are processed daily by an EMR job.
Recently it has been observed that the file sizes vary, and the EMR jobs take too long. The distributor needs to tune and optimize the data processing workflow with this limited information to improve the performance of the
EMR job.
Which recommendation should an administrator provide?

A. Reduce the HDFS block size to increase the number of task processors.

B. Use bzip2 or Snappy rather than gzip for the archives.

C. Decompress the gzip archives and store the data as CSV files.

D. Use Avro rather than gzip for the archives.