Analytics services

One pipeline: collect, store, prepare, query, show

CollectStore and prepareQueryShowClicks, logs,sensorsAmazon Kinesisstreams data inas it happensAmazon S3the data lakeAWS Gluecatalogues and preparesAmazon AthenaSQL straight on the S3 filesAmazon Redshiftdata warehouseAmazon EMRSpark and Hadoop clustersAmazon Quickdashboards(was QuickSight)The analytics services are stages of one pipeline: collect, store, prepare, query, show.Place each service on the pipeline and the questions answer themselves. Kinesis is the only one about dataarriving in real time. Glue is the only one that prepares data (ETL). Athena needs no servers and no loading;Redshift is a warehouse you load; EMR is for teams that already use Spark or Hadoop.

Kinesis collects, S3 stores, Glue prepares, Athena, Redshift and EMR query, Quick shows.

Analytics services · Going deeper

Clue words for analytics

When the question saysThe service isBecause it"SQL on files in S3", "serverless queries"Amazon Athenapay for the data each query scans"discover, catalogue, transform", "ETL"AWS Glueserverless data preparation"Spark", "Hadoop", "big data frameworks"Amazon EMRmanaged clusters"real-time", "streaming data"Amazon Kinesisingests streams as they arrive"data warehouse", "petabytes", "complex SQL"Amazon Redshiftcolumnar warehouse"dashboards", "business intelligence"Amazon QuickBI service, formerly QuickSight"search", "log analytics"Amazon OpenSearch Servicemanaged search engineSeven analytics services. "Real-time" and "ETL" each point to exactly one of them.

Seven analytics services, seven clue phrases.

AI and machine learning

Three layers: use a model, call a model, build a model

You bringAI servicesready-made models, one task each: Rekognition, Transcribe, Polly,Translate, Comprehend, Textract, Lex, Kendraa call to an APIAmazon Bedrockfoundation models from several companies behind one API, forgenerative AI; Amazon Q is an assistant built this waya prompt and your dataAmazon SageMaker AIthe workshop for building, training and deployingmachine learning models of your owndata, code and ML skillUse a ready-made service, call a foundation model, or build your own model.Going down, you get more control and need more skill. For one clear task such as reading text from a scan, anAI service is the answer. For generative AI, the answer is Bedrock. SageMaker AI is the answer only when ateam wants to train or tune its own model.

AI services do one task. Bedrock is generative AI. SageMaker AI builds your own.

AI and machine learning · Going deeper

Clue words for AI services

When the question saysThe service isBecause it"find objects or faces in images and video"Amazon Rekognitioncomputer vision"turn speech into text"Amazon Transcribespeech recognition"turn text into lifelike speech"Amazon Pollytext to speech"translate between languages"Amazon Translatemachine translation"find sentiment or key phrases in text"Amazon Comprehendnatural language processing"extract text and tables from scanned documents"Amazon Textractdocument reading"build a chatbot or voice bot"Amazon Lexconversational interfaces"search across company documents"Amazon Kendraintelligent enterprise searchTranscribe hears, Polly speaks, Comprehend reads meaning, Textract reads documents.

Eight AI services, each for one task.

Check yourself

scanned form?name: ...date: ...amount: ...

An insurer receives scanned claim forms and needs the typed and handwritten text taken out of them. Which service?

Application integration

A queue holds it for one; a topic sends it to all

A queueone receiver, laterA topicmany receivers, nowOrderssendsAmazon SQSholds each messagetakes one when readyBillingOrderspublishesAmazon SNSpushes at oncean email addressa text messagea Lambda functionSQS holds a message until one consumer takes it. SNS pushes it to every subscriber at once.A queue decouples two systems: if billing is slow or down, orders wait safely. A topic fans out: one event,many listeners. Two more services sit beside them: Amazon EventBridge routes events from AWS services byrule or on a schedule, and AWS Step Functions runs the steps of a workflow in order.

SQS is a queue that decouples. SNS is a topic that fans out.

Application integration · Going deeper

Clue words for integration

When the question saysThe service isBecause it"decouple", "buffer", "no message lost"Amazon SQSmessage queue"notify", "fan out", "email or text alerts"Amazon SNSpublish and subscribe"react to events from AWS services", "on a schedule"Amazon EventBridgeevent bus with rules"coordinate the steps of a workflow"AWS Step Functionsvisual workflow orchestrationFour ways for parts of an application to talk without calling each other directly.The exam calls this "loose coupling": each part can fail, scale or change without breaking the others.

SQS, SNS, EventBridge and Step Functions.

Check yourself

Orders?EmailStockAnalytics

When an order is placed, the email, stock and analytics systems must each be told at the same moment. Which service fits?

Developer tools

Build, deploy, and the pipeline that runs both

Your codein a repositoryAWS CodePipeline: runs the stages in order on every changeAWS CodeBuildcompiles and testsAWS CodeDeployinstalls it on serversProductionAWS X-Ray traces each requestthrough the running applicationAWS CLI, SDKs and CloudShellcall AWS from commands and codeBuild, deploy, pipeline: each developer tool is named for its stage.CodeBuild builds and tests. CodeDeploy puts the result on servers. CodePipeline is the conveyor that runs thestages automatically on every change. X-Ray is for after release: it shows where a request spent its time.

CodeBuild builds, CodeDeploy deploys, CodePipeline automates, X-Ray traces.

Other categories

One line each is all the exam asks

When the question saysThe service isIts category"a contact centre in the cloud"Amazon Connectbusiness application"send marketing and transactional email"Amazon SESbusiness application"virtual desktops for employees"Amazon WorkSpacesend-user computing"safe access to internal sites from a browser"WorkSpaces Secure Browserend-user computing"build and host a web or mobile app quickly"AWS Amplifyfrontend web and mobile"one GraphQL API over several data sources"AWS AppSyncfrontend web and mobile"connect and manage devices"AWS IoT Coreinternet of thingsFor these, the exam asks only what the service is for.Task 3.8 covers the categories outside compute, storage, networking and databases. One line each is enough:no question goes deeper than matching the need to the name.

Match the need to the name: Connect, SES, WorkSpaces, Amplify, AppSync, IoT Core.

Check yourself

at home?company desktop

Staff working from home need a full Windows desktop that the company manages centrally. Which service provides it?

Lesson complete

You finished “AI, analytics and the rest of the catalogue”

Answer the questions along the way to see your score here.

Open in a new tab