AWS Databases and Analytics Services Overview – SAP-C02 Study Guide

Purpose of This Lesson

AWS database and analytics services appear frequently in architecture exam scenarios because workloads have very different requirements for data models, performance, scalability, consistency, processing, and cost. The first step is recognizing the category of service that matches the workload.

This lesson introduces the major AWS services covered in the databases and analytics domain:

• Relational databases: Amazon RDS and Amazon Aurora
• NoSQL databases: Amazon DynamoDB
• Data warehouses: Amazon Redshift
• Big data processing: Amazon EMR
• Streaming data ingestion and delivery: Amazon Kinesis Data Streams and Amazon Kinesis Data Firehose

Later lessons typically examine the capabilities and tradeoffs of each service in greater depth. For exam questions, service selection is usually driven by workload characteristics rather than by a general preference for one database technology.

Key Concepts

Relational database workloads

Relational databases are appropriate when an application requires structured tables, relationships between entities, SQL queries, transactions, and strong consistency controls. In AWS, the main managed relational options include:

• Amazon RDS: Managed relational database instances supporting multiple database engines.
• Amazon Aurora: AWS-designed relational database compatible with commonly used MySQL and PostgreSQL workloads, with a cloud-oriented storage architecture.

Typical use cases include transactional applications, enterprise systems, order processing, and applications that depend on SQL and relational integrity.

NoSQL database workloads

NoSQL services are designed for highly scalable applications that can work with nonrelational data models and access patterns. Amazon DynamoDB is a managed key-value and document database that can provide low-latency access at scale.

DynamoDB design is strongly driven by the application’s access patterns. A successful design generally identifies the keys and queries first, then models the table around those requirements.

Data warehousing

A data warehouse is optimized for analytical queries over large datasets rather than high-volume transactional updates. Amazon Redshift is AWS’s managed data warehouse service and is intended for reporting, business intelligence, and complex analytical workloads.

A common architectural distinction is:

• Online transaction processing (OLTP): Frequent inserts, updates, and point lookups; commonly relational or NoSQL.
• Online analytical processing (OLAP): Large scans, aggregations, and reporting; commonly a data warehouse such as Amazon Redshift.

Big data processing

Amazon EMR is a managed platform for running distributed data-processing frameworks. It is useful when a workload requires large-scale processing with technologies such as Hadoop or Spark, or when teams need more control over the processing environment than a fully abstracted analytics service provides.

EMR is a processing platform, not simply a general-purpose database. Its role is to transform, analyze, or process large datasets, often stored in services such as Amazon S3.

Streaming data services

Streaming architectures handle data that arrives continuously, such as application events, telemetry, logs, or clickstreams.

• Amazon Kinesis Data Streams: Captures and retains streaming records for applications or consumers to process. It is appropriate when consumers need control over how and when records are read and processed.
• Amazon Kinesis Data Firehose: Delivers streaming data to supported destinations with a managed delivery model. It is useful when the primary requirement is reliable ingestion and delivery rather than custom stream-consumer processing.

A key exam distinction is whether the architecture requires a durable stream that multiple consumers process independently, or a simpler managed delivery pipeline into a destination.

Architecture Decision Guide

Requirement	Likely service category	AWS service to evaluate	Main decision factor
SQL transactions and relational data	Relational database	Amazon RDS or Amazon Aurora	Database engine compatibility, availability, performance, and operational requirements
Highly scalable key-value or document access	NoSQL database	Amazon DynamoDB	Access patterns, key design, latency, and scaling requirements
Complex reporting and large analytical queries	Data warehouse	Amazon Redshift	Query performance, warehouse architecture, and analytical data volume
Distributed processing using big data frameworks	Big data processing	Amazon EMR	Framework support and required cluster or runtime control
Continuous event ingestion with multiple consumers	Streaming platform	Amazon Kinesis Data Streams	Consumer control, retention, ordering, and stream-processing requirements
Managed delivery of streaming records to a destination	Streaming delivery	Amazon Kinesis Data Firehose	Minimal operational management and supported delivery targets

Exam-Relevant Takeaways

• Select the database based on the data model and access pattern, not merely the amount of data.
• Relational databases are commonly associated with SQL, transactions, relationships, and structured schemas.
• DynamoDB is a managed NoSQL service designed for predictable, low-latency access at scale; its table design must reflect application queries.
• Amazon Redshift is intended for analytical workloads and should not automatically be selected as the primary transactional database.
• Amazon EMR provides distributed data-processing capabilities and is distinct from a database or data warehouse.
• Kinesis Data Streams and Kinesis Data Firehose solve related but different problems: stream processing control versus managed delivery.
• Architecture questions may combine services. For example, operational data can be exported or transformed before being analyzed in a data warehouse, while streaming events can be delivered to storage or analytics destinations.
• Consider operational burden as well as technical fit. Managed services reduce infrastructure administration, but they may impose service-specific data-model, integration, or processing constraints.

Common Exam Traps

• Choosing a relational database solely because the workload is important: Importance does not determine the data model. A high-scale workload may be better suited to DynamoDB or another specialized service.
• Confusing Redshift with a transactional database: Redshift is optimized for analytics, aggregations, and reporting, not general-purpose OLTP workloads.
• Treating EMR as a database: EMR runs distributed processing frameworks; it is not a replacement for a persistent transactional or analytical database.
• Confusing Kinesis Data Streams with Firehose: Data Streams supports custom consumers and stream-processing logic. Firehose focuses on managed delivery to destinations.
• Ignoring query patterns in DynamoDB: A flexible or relational-style schema does not automatically translate into a good DynamoDB design. Access patterns and key design are central.
• Assuming every database must be self-managed on Amazon EC2: AWS managed database services are usually preferred when they satisfy the requirements and reduce operational responsibility.

Real-World Engineer Notes

Start database selection by documenting the workload rather than listing preferred services. Useful questions include:

1. Is the workload transactional, analytical, or stream-processing oriented?
2. Does it require SQL joins and relational constraints?
3. What are the dominant read and write patterns?
4. Is latency predictable and low, or are large scans and aggregations more important?
5. Does the processing layer require a particular framework or custom consumer behavior?
6. Which operational tasks should the platform team manage, and which can AWS manage through a higher-level service?

It is common for a production architecture to use several categories together. For example, a transactional application may use a relational database, publish events to a streaming service, store historical data in object storage, and load curated data into a warehouse for reporting. The services are complementary rather than interchangeable.

Quick Reference Summary

• Amazon RDS: Managed relational database service supporting multiple engines.
• Amazon Aurora: AWS-designed relational database compatible with MySQL and PostgreSQL workloads.
• Amazon DynamoDB: Managed key-value and document NoSQL database for scalable, low-latency access.
• Amazon Redshift: Managed data warehouse for analytical queries and reporting.
• Amazon EMR: Managed platform for distributed big data processing frameworks.
• Kinesis Data Streams: Durable streaming ingestion with application-controlled consumers.
• Kinesis Data Firehose: Managed streaming data delivery to supported destinations.
• Primary selection rule: Match the service to the workload’s data model, access pattern, processing model, and operational requirements.

Flashcards

1. Q: Which AWS services represent the main managed relational database options introduced in this lesson?

A: Amazon RDS and Amazon Aurora.

1. Q: What type of database is Amazon DynamoDB?

A: A managed NoSQL key-value and document database.

1. Q: What workload is Amazon Redshift designed to support?

A: Data warehousing and analytical workloads involving reporting, aggregations, and large queries.

1. Q: What is the primary role of Amazon EMR?

A: Running distributed big data processing frameworks such as Hadoop or Spark.

1. Q: What is the key architectural distinction between Kinesis Data Streams and Kinesis Data Firehose?

A: Data Streams provides application-controlled stream consumption, while Firehose provides managed delivery to supported destinations.

1. Q: What should drive the design of a DynamoDB table?

A: The application’s known access patterns, including the keys and queries it must support.

1. Q: Which workload is generally associated with OLTP: transactional updates or large analytical scans?

A: Transactional updates.

1. Q: Which workload is generally associated with OLAP?

A: Large analytical scans, aggregations, and reporting.

1. Q: Why is Amazon Redshift usually not selected as an application’s primary transactional database?

A: It is optimized for analytical processing rather than general-purpose transactional access.

1. Q: Are EMR, Redshift, and DynamoDB interchangeable?

A: No. EMR is a processing platform, Redshift is a data warehouse, and DynamoDB is a NoSQL operational database.

Practice Questions

Question 1

A company is building an order-processing application. It requires SQL queries, transactions, relationships between customers and orders, and a managed database service. Which category should the architect evaluate first?

A. Amazon Kinesis Data Firehose
B. Amazon Redshift
C. Amazon RDS or Amazon Aurora
D. Amazon EMR

Correct answer: C

Explanation: The workload is a structured, transactional OLTP application with relational requirements. Amazon RDS or Amazon Aurora should be evaluated before analytical or streaming services.

Question 2

A web application must support very high request volumes with predictable low-latency access to user profiles. The application primarily retrieves and updates records by a known user identifier and does not require relational joins. Which service is the best initial choice?

A. Amazon DynamoDB
B. Amazon Redshift
C. Amazon EMR
D. Amazon Kinesis Data Firehose

Correct answer: A

Explanation: DynamoDB is designed for scalable key-value and document access with low latency. The known identifier provides a natural basis for the table’s access-pattern-driven key design.

Question 3

A company needs to collect application events continuously. Several independent processing applications must read the same events and control their own processing position. Which service is the best fit?

A. Amazon Kinesis Data Firehose
B. Amazon Kinesis Data Streams
C. Amazon Redshift
D. Amazon Aurora

Correct answer: B

Explanation: Kinesis Data Streams is intended for application-controlled consumption and supports multiple consumers processing a stream independently. Firehose is better suited to managed delivery when custom consumer control is not required.

Question 4

An organization wants to run distributed Spark processing over a large dataset and requires control over the processing environment and framework configuration. Which service should it evaluate?

A. Amazon EMR
B. Amazon DynamoDB
C. Amazon RDS
D. Amazon Kinesis Data Firehose

Correct answer: A

Explanation: Amazon EMR provides a managed platform for distributed big data frameworks such as Spark and is appropriate when the workload needs framework and runtime control.

Question 5

A business intelligence team needs to run complex aggregations and reporting queries over a large historical dataset. The workload is analytical rather than transactional. Which service category is most appropriate?

A. NoSQL operational database
B. Data warehouse
C. Relational OLTP database only
D. Stream ingestion service only

Correct answer: B

Explanation: Data warehouses such as Amazon Redshift are designed for analytical queries, aggregations, and reporting over large datasets. A transactional database or stream ingestion service alone does not directly address this requirement.