Domain 3 review: databases, networking and other services

The second half of the Domain 3 review covers the services that store structured data, connect everything together, and add analytics, artificial intelligence and integration on top. Kwame is more at home here than Mateo: BookVet talks to a database, sends confirmation emails and will soon need a queue for reminder messages. Gwen asks him to explain to Mateo what the booking app does when a client books an appointment, and then names the AWS service that would handle each step. By the end of the walkthrough the whiteboard has a dozen service names on it, and Mateo has a page of one-line definitions in his study log.

Databases

Task statement 3.4 is about choosing database services. A relational database stores data in tables with rows and columns and uses SQL; it suits structured data with relationships, such as clients, pets and appointments. Amazon RDS is a managed relational database service that supports several engines, including MySQL, PostgreSQL, MariaDB, Oracle and Microsoft SQL Server, and handles tasks such as backups, patching and Multi-AZ standby copies. Amazon Aurora is AWS's own relational engine, compatible with MySQL and PostgreSQL, designed for high performance and availability. Running a database yourself on EC2 gives full control but means you manage everything: installation, patching, backups and high availability. The exam often contrasts these: a question asking to reduce administrative effort for a relational database points to RDS or Aurora rather than EC2.

Non-relational, or NoSQL, databases suit other data shapes. Amazon DynamoDB is a fully managed key-value and document database that delivers single-digit millisecond performance at any scale and is serverless. Amazon ElastiCache and Amazon MemoryDB provide in-memory data stores for caching and very fast access. Amazon DocumentDB stores JSON-style documents and is compatible with MongoDB. Amazon Neptune is a graph database for highly connected data, such as social networks or fraud detection. For migrations, AWS Database Migration Service moves data between databases with minimal downtime, and the AWS Schema Conversion Tool converts schemas when changing engines.

Networking and content delivery

Amazon Virtual Private Cloud, Amazon VPC, gives you a logically isolated network in AWS. A VPC is divided into subnets, each in one Availability Zone. Public subnets have a route to an internet gateway, which connects the VPC to the internet; private subnets do not. A NAT gateway lets resources in private subnets reach the internet for updates without accepting incoming connections. Route tables decide where traffic goes. Security groups and network ACLs, reviewed in lesson 4, control access. Cascade Ridge's plan puts the BookVet load balancer in public subnets and the application servers and RDS database in private subnets, the standard layered design.

Several services connect AWS to the outside world. Amazon Route 53 is AWS's domain name system service, which translates names such as the booking site's address into IP addresses and can route users based on health or location. Amazon CloudFront is a content delivery network that caches content at edge locations close to users. AWS Global Accelerator improves the performance of applications for global users by routing traffic over the AWS network. For connections to on-premises networks, AWS Site-to-Site VPN creates an encrypted tunnel over the internet, which is quick to set up, while AWS Direct Connect provides a dedicated private connection with more consistent performance, which takes longer to arrange. AWS Transit Gateway connects many VPCs and on-premises networks through a central hub. AWS Client VPN gives individual users secure remote access.

Analytics

Task statement 3.7 covers analytics services. Amazon Athena runs SQL queries directly on data in S3 without loading it into a database, and you pay per query. AWS Glue discovers, prepares and transforms data, and keeps a data catalog. Amazon Redshift is a data warehouse for fast analytics on large structured data sets. Amazon EMR runs big data frameworks such as Apache Spark. Amazon Kinesis collects and processes streaming data in real time, such as clickstreams or sensor readings. Amazon OpenSearch Service supports search and log analytics. Amazon QuickSight builds dashboards and visualizations; AWS now offers it as part of the broader Amazon Quick Suite, so you may see either name. Lorraine wants a monthly dashboard of appointments and revenue by clinic; Gwen sketches Glue to prepare the exports, Athena or Redshift to query them, and QuickSight for the dashboard.

Artificial intelligence and machine learning

AWS offers AI at three levels, and the exam asks you to recognize which level fits a need. At the top are ready-made AI services that solve one task through an API, with no machine learning expertise needed. Amazon Rekognition analyses images and video. Amazon Comprehend finds meaning in text, such as sentiment or key phrases. Amazon Transcribe turns speech into text, and Amazon Polly turns text into speech. Amazon Translate translates text. Amazon Textract extracts text and data from documents and forms. Amazon Lex builds chatbots. Amazon Kendra provides intelligent search across documents. In the middle, Amazon Bedrock gives access to foundation models from several providers for generative AI applications, and Amazon Q provides generative AI assistants for business and development work. At the base, Amazon SageMaker AI lets data scientists build, train and deploy their own machine learning models.

A practice question might describe the reception team wanting phone calls transcribed so they can search them later: that is Transcribe, perhaps with Comprehend to find topics. Scanning paper intake forms from new clients into the practice system is Textract. A chatbot on the booking site that answers questions about opening hours is Lex, or a generative AI assistant built with Bedrock. Training a custom model to predict which appointments are likely to be missed, using the clinic group's own history, is SageMaker AI.

Other in-scope service families

Task statement 3.8 lists several more categories. Application integration services connect parts of an application. Amazon Simple Queue Service, SQS, is a message queue that decouples components so one part can work at its own pace; BookVet's reminder messages can wait in a queue until a worker sends them. Amazon Simple Notification Service, SNS, sends messages to many subscribers at once, such as email or text alerts. Amazon EventBridge routes events between services and applications. AWS Step Functions coordinates multi-step workflows. Business applications include Amazon Connect, a cloud contact center, and Amazon Simple Email Service, SES, for sending email such as booking confirmations.

Developer tools include AWS CodeBuild, which compiles and tests code, AWS CodePipeline, which automates release pipelines, and AWS X-Ray, which traces requests through an application to find problems. End-user computing includes Amazon WorkSpaces, which provides virtual desktops, and Amazon AppStream 2.0, which streams applications to a browser. Front-end and mobile services include AWS Amplify for building and hosting web and mobile apps. Internet of Things services include AWS IoT Core for connecting devices; Mateo jokes that the kennel temperature sensors could report there. Management and governance services include AWS CloudFormation, AWS Config, AWS Systems Manager for managing fleets of servers, AWS Organizations for managing multiple accounts, and AWS Health Dashboard for service events that affect your account.

Managed versus self-managed: the thread through Domain 3

One idea runs through almost every Domain 3 question about databases, analytics and integration: how much does the customer have to manage? Gwen draws a sliding scale on the whiteboard. At one end is running software yourself on EC2: you install the database, patch it, back it up, configure replication and plan for failure. In the middle are managed services such as Amazon RDS, where AWS handles installation, patching, backups and standby copies, and you handle the schema, the data and access. At the far end are serverless services such as DynamoDB, Athena, Lambda and SQS, where there are no servers to size or patch at all and you pay for what you use.

Questions that include phrases such as "least operational overhead", "minimize administration" or "the team has little database experience" are pointing along that scale toward managed or serverless options. Questions that mention a need for full control of the operating system, a database engine version AWS does not offer, or software that must be installed in a particular way are pointing back toward EC2. Neither end is always right. The exam rewards matching the option to the stated priority.

Kwame uses the scale to sort the BookVet roadmap. Phase 1 keeps the app on EC2 because it is the fastest way to move it unchanged. Phase 2 moves the booking logic to Lambda, the appointment records to DynamoDB, the reminder queue to SQS and the confirmation emails to SES, sliding each part toward the serverless end. The application does the same job, but there are far fewer servers for him to look after at three in the morning, and the bill follows the number of bookings rather than the size of the servers.

Mateo's one-line definitions page ends up being one of his most useful study tools. He reads it on the train and quizzes himself by covering the right-hand column, a habit that lesson 8 builds into a full method for confused pairs.