Developer tools, end-user computing, front end and IoT

Learn the developer tools by their verb

AWS CodePipeline decides the ORDER: which stages exist, what runs next, and the condition that one must succeed before the next begins. AWS CodeBuild MAKES the artifact — it compiles and runs the tests. AWS CodeDeploy PLACES an artifact that already exists onto EC2 instances, Lambda or ECS, with rollback. The source is AWS CodeCommit or a third-party repository the pipeline connects to. A question describing a sequence, visible and repeatable rather than living in somebody's shell history, is describing the first of those; either of the others alone is one stage of it.

Around them sit the tools that show up as plausible distractors. AWS Cloud9 is a browser-based development environment and AWS CloudShell is a browser shell with the CLI already in it. AWS X-Ray traces one request across the services it touches, which is a debugging tool rather than a delivery one. AWS Amplify builds and hosts web and mobile front ends, with AWS AppSync behind managed APIs. And AWS CloudFormation is very often a STEP INSIDE a pipeline — it provisions infrastructure from a template — but it never decides which step runs next.

Desktops and applications delivered to people

Amazon WorkSpaces provides managed virtual desktops that people connect to from whatever device they own, billed per user with monthly or hourly options that suit a workforce which arrives and leaves. Amazon AppStream 2.0 streams a single application to a browser session instead. Both keep the data in AWS rather than on the endpoint, and that — rather than cost — is usually the real reason an organisation buys either: a lost contractor laptop then contains nothing.

Because both satisfy the "own laptop, no company data on it" clause, that clause never decides the answer; the scope does. Two hundred seasonal contractors needing a full Windows desktop with the company's applications is WorkSpaces. Building the same thing out of EC2 Windows instances and RDP is possible and hands you the images, the provisioning, the patching, the licensing and the directory integration for people who turn over every year, which is the work the service exists to remove.

Devices that must keep working when the link does not

AWS IoT Core is the front door in the cloud: device identity and registry, MQTT messaging, a device shadow holding each device's last known and desired state, and a rules engine that routes messages onward to other services. Everything it does runs in the Region, which is the fact the exam builds its question around.

AWS IoT Greengrass takes that programming model out to hardware in the building, so messaging, application logic and even machine learning inference run locally and keep running when the connection to AWS is gone, synchronising when it returns. A refrigeration unit that must close a valve within a second of a bad reading — including while the shop's internet is down — is deciding locally, and only an edge runtime does that. It is the hybrid deployment model in miniature: the part that must survive without the cloud stays where the physical thing is.