AWS VI: Architecting on AWS & the exams: glossary
124 terms, in plain words, with where each one is taught.
- 7 Rs (seven Rs, migration strategies, migration strategy)
- The seven ways to migrate one application: retire, retain, rehost, relocate, repurchase, replatform, refactor. Lesson 39.36
- Amazon API Gateway (API Gateway, aws apigateway, aws apigatewayv2)
- The managed front door for APIs: HTTPS, routing to Lambda or HTTP backends, throttling, authorization, logging and its own error codes. Lesson 39.12
- Amazon Athena (Athena)
- Serverless SQL over files in S3, billed by data scanned: columnar formats and partitions make queries cheaper and faster. Lesson 39.23
- Amazon Cognito (Cognito, user pool, identity pool)
- Customer identity: user pools for sign-up, sign-in, MFA and tokens; identity pools to trade a token for temporary AWS credentials. Lesson 39.29
- Amazon Data Firehose (Firehose, Kinesis Data Firehose, delivery stream)
- Buffered, near-real-time delivery of records to S3, Redshift, OpenSearch or Splunk with no consumer code, optional format conversion and transformation. Lesson 39.23
- Amazon ECR (ECR, Elastic Container Registry, aws ecr)
- AWS's private container image registry: repositories with scanning, lifecycle policies and immutable tags; Docker logs in with a 12-hour token. Lesson 39.15
- Amazon ECS (ECS, Elastic Container Service, aws ecs)
- AWS's own container orchestrator: task definitions, tasks and services in clusters, on Fargate or on EC2 instances. Lesson 39.15
- Amazon ElastiCache (ElastiCache, Valkey, Memcached)
- Managed in-memory caches: Valkey or Redis OSS (replication, persistence, data structures) and Memcached (simple, multi-threaded). Lesson 39.26
- Amazon EMR (EMR, EMR Serverless)
- Managed big-data frameworks - Spark, Hive, Presto - on clusters of EC2, on EKS, or serverless. Lesson 39.23
- Amazon MQ (ActiveMQ, RabbitMQ)
- A managed ActiveMQ / RabbitMQ broker for applications that already speak JMS, AMQP or MQTT and move to AWS without rewriting their messaging. Lesson 39.1
- Amazon SQS queue (SQS queue, standard queue, aws sqs)
- A managed message queue: producers send, consumers receive and delete. Standard queues are nearly unlimited in throughput and deliver at least once in roughly the order sent. Lesson 39.1
- Amazon States Language (ASL, States Language)
- The JSON in which a state machine is written: StartAt, States, and states of type Task, Choice, Wait, Parallel, Map, Pass, Succeed and Fail. Lesson 39.20
- Application Auto Scaling (register-scalable-target, put-scaling-policy, TargetTrackingScaling)
- Scaling for ECS services and other resources: a scalable target (min / max) and a target tracking policy that keeps a metric near a value with its own alarms. Lesson 39.15
- asynchronous invocation (--invocation-type Event, async invocation, Event invocation)
- Lambda queues the event, answers 202 at once and runs it; on errors it retries twice, then sends the event to a destination or DLQ if configured. Lesson 39.5
- AWS App Runner (App Runner, Elastic Beanstalk, AWS Elastic Beanstalk)
- Services that hide even the cluster and load balancer for simple web apps: App Runner runs an image or repository as a web service; Elastic Beanstalk deploys an app onto managed EC2. Lesson 39.15
- AWS Application Migration Service (Application Migration Service, AWS MGN, MGN)
- The rehost service: an agent replicates a source server's disks continuously into AWS, test instances launch from the copy, and the cutover takes minutes. Lesson 39.36
- AWS Cloud Adoption Framework (AWS CAF, Cloud Adoption Framework, CAF)
- AWS's guidance for the organisational side of moving to the cloud: six perspectives of capabilities to build, four business outcomes (reduced business risk, better ESG performance, more revenue, more operational efficiency) and the phases envision, align, launch, scale. Lesson 39.36
- AWS Control Tower (Control Tower, landing zone, Account Factory)
- Sets up and governs a multi-account landing zone on AWS Organizations: log archive and audit accounts, guardrails (SCPs, Config rules) and Account Factory. Lesson 39.29
- AWS Data Transfer Terminal (Data Transfer Terminal)
- AWS sites where you bring your own storage devices and upload them to AWS at high speed: the offline path now that Snowball Edge is closed to new customers. Lesson 39.36
- AWS DataSync (DataSync, AWS Transfer Family, Transfer Family)
- Online transfer services: DataSync copies files between on-premises storage and S3 / EFS / FSx on a schedule; Transfer Family offers SFTP / FTPS / FTP endpoints into S3. Lesson 39.23
- AWS Fargate (Fargate, FARGATE launch type)
- Serverless capacity for containers: each ECS (or EKS) task runs on capacity AWS manages, with its own network interface, and is billed per vCPU and GB per second. Lesson 39.15
- AWS Global Accelerator (Global Accelerator)
- Static anycast IP addresses that route TCP and UDP traffic over the AWS network to the nearest healthy Regional endpoint - not a cache. Lesson 39.26
- AWS Glue (Glue, Glue Data Catalog, Glue crawler)
- Serverless data integration: crawlers fill the Data Catalog with table definitions, and ETL jobs transform data, for example CSV to Parquet. Lesson 39.23
- AWS Lake Formation (Lake Formation)
- Builds and secures data lakes on S3 and the Glue catalog, with table-, column- and row-level permissions. Lesson 39.23
- AWS Lambda (Lambda function, aws lambda, serverless function)
- Run code without managing servers: a function runs when invoked, scales by itself and is billed per request and per GB-second of run time. Lesson 39.5
- AWS Lambda Power Tuning (Power Tuning)
- An open-source Step Functions tool that runs a function at several memory sizes and shows the duration and cost of each, to pick the cheapest or fastest. Lesson 39.5
- AWS Pricing Calculator (Pricing Calculator)
- AWS's web tool to estimate the monthly cost of an architecture before building it. Lesson 39.31
- AWS Step Functions (Step Functions, state machine, aws stepfunctions)
- A managed workflow service: a state machine in JSON whose steps call functions and AWS services, with declared retries and error handling and a recorded history. Lesson 39.20
- AWS Storage Gateway (Storage Gateway, file gateway, AWS Snowball)
- Hybrid storage that gives on-premises applications file, volume or tape interfaces backed by AWS with a local cache; Snowball moved terabytes offline on a shipped device (closed to new customers since 7 November 2025). Lesson 39.23
- AWS Transform (AWS Migration Hub, Migration Hub, AWS Application Discovery Service)
- AWS's migration and modernization service (2025): discovery, assessment, wave planning and modernization agents. It replaces Migration Hub and Application Discovery Service, closed to new customers on 7 November 2025. Lesson 39.36
- backup and restore (backup-restore)
- The cheapest DR strategy: only backups exist in the recovery Region and everything is rebuilt from code; RPO and RTO in hours. Lesson 39.27
- buffering hints (BufferingHints, SizeInMBs, IntervalInSeconds)
- Firehose's size and interval: an object is written when either is reached. They trade latency against the number and size of objects. Lesson 39.23
- cache stampede (thundering herd, dog-pile)
- Many requests missing the same expired key at once and all hitting the database together. Lesson 39.26
- cache-aside (lazy loading, write-through, write-through cache)
- Cache-aside: read the cache, on a miss read the database and fill the cache with a TTL. Write-through: update the cache on every write. Lesson 39.26
- CAF perspectives (Business perspective, People perspective, Governance perspective)
- The six groups of capabilities in AWS CAF: Business, People and Governance (business capabilities), Platform, Security and Operations (technical capabilities). Lesson 39.36
- Catch (Step Functions) (Catch, States.ALL, ResultPath)
- A state's list of errors that go to another state (with the error stored under ResultPath) - for permanent failures, compensation and clean endings. Lesson 39.20
- change data capture (CDC, ongoing replication)
- Copying every change a source database makes after the initial full load, so the copy stays current until the application switches over. Lesson 39.36
- CLF-C02 (Cloud Practitioner, AWS Certified Cloud Practitioner)
- The foundational AWS exam: 65 questions, 90 minutes, 700 to pass, on what services are for, security, pricing and support. Lesson 39.38
- cold start (cold starts, INIT_START, Init Duration)
- The first invocation in a new execution environment: the code is loaded, the runtime started and the module's top-level code run. Shown as Init Duration in the REPORT line. Lesson 39.5
- concurrency (Lambda) (Lambda concurrency, ConcurrentExecutions, concurrent executions)
- How many invocations of a function run at the same moment - about request rate x duration. An account has a Regional pool, 1,000 by default. Lesson 39.5
- confused deputy (confused deputy problem, aws:SourceArn, aws:SourceAccount)
- A service tricked into using the permission you granted it for someone else's resource; prevented by pinning resource policies to the source with aws:SourceArn and aws:SourceAccount. Lesson 39.29
- data lake (data lakes)
- Data of many kinds kept in S3 in open formats, cataloged and queried by several engines (Athena, EMR, Redshift Spectrum) instead of one database. Lesson 39.23
- dead-letter queue (SQS) (DLQ, dead-letter queue, dead letter queue)
- A queue where messages that failed too often are moved by the source queue's redrive policy, so they stop looping and can be inspected and redriven later. Lesson 39.1
- decoupling (loose coupling, loosely coupled, decouple)
- Putting a queue, topic or event bus between components so a producer hands work over instead of calling the consumer directly; a slow or failing consumer then builds a backlog instead of an outage. Lesson 39.1
- Dedicated Hosts (Dedicated Host, Dedicated Instances, BYOL)
- A physical server dedicated to you, for per-socket or per-core licences and compliance; Dedicated Instances only guarantee no other customer's instances on the hardware. Lesson 39.31
- deduplication ID (MessageDeduplicationId, ContentBasedDeduplication, content-based deduplication)
- How a FIFO queue recognises a repeated send within five minutes: an explicit id, or a hash of the body with content-based deduplication. Lesson 39.1
- deployment circuit breaker (deploymentCircuitBreaker, circuit breaker (ECS))
- An ECS service setting that stops a rolling deployment whose tasks keep failing and, with rollback, returns to the last deployment that completed. Lesson 39.15
- disaster recovery (DR, DR strategy, DR strategies)
- Restoring service after losing a Region or after a mistake that replication copies, from another Region or from backups, within an agreed RPO and RTO. Lesson 39.27
- DynamoDB Accelerator (DAX)
- An in-memory cache in front of DynamoDB that answers reads in microseconds through a DynamoDB-compatible client. Lesson 39.26
- ECS service (create-service, desiredCount, ECS services)
- Keeps a desired number of tasks running, replaces failed ones, registers them in a load balancer target group and rolls out new revisions. Lesson 39.15
- Endpoint request timed out (504 (API Gateway), integration timeout)
- The 504 API Gateway returns when the integration does not answer within its timeout (29 seconds by default for REST APIs, 30 for HTTP APIs). Lesson 39.12
- event source mapping (ESM, create-event-source-mapping, event source mappings)
- Lambda polling a queue or stream for you (SQS, Kinesis, DynamoDB Streams, Kafka, MQ) and invoking the function with batches of records. Lesson 39.6
- execution environment (execution environments, warm environment, warm start)
- The isolated runtime that runs a function's invocations, one at a time, reused while it stays warm and recycled after idle time, errors or updates. Lesson 39.5
- execution role (Lambda) (Lambda execution role, AWSLambdaBasicExecutionRole, AWSLambdaSQSQueueExecutionRole)
- The IAM role a function runs as: its code's AWS calls, writing its logs and polling its event sources all use it. It must trust lambda.amazonaws.com. Lesson 39.5
- Express workflow (Standard workflow, EXPRESS, start-sync-execution)
- Step Functions' high-volume, short workflow type (up to 5 minutes, at least once). Standard workflows run up to a year, exactly once, with a full history. Lesson 39.20
- fan-out (SNS fan-out, fanout)
- Publishing one message to an SNS topic so every subscriber gets its own copy - typically one SQS queue per consumer, so each consumer keeps its own backlog and retries. Lesson 39.1
- Fargate Spot (FARGATE_SPOT, capacity provider strategy, capacity provider)
- Spare Fargate capacity at a large discount that can be reclaimed with a two-minute warning, chosen per service through a capacity provider strategy. Lesson 39.15
- FIFO queue (.fifo, FifoQueue, SQS FIFO)
- An SQS queue (name ending in .fifo) that keeps order per message group and drops duplicates within five minutes, at lower throughput than a standard queue. Lesson 39.1
- filter policy (FilterPolicy, FilterPolicyScope, subscription filter policy)
- A JSON policy on an SNS subscription that decides which messages it receives, matching message attributes by default or the body with FilterPolicyScope MessageBody. Matching is typed. Lesson 39.1
- function URL (Lambda function URL, lambda-url)
- A dedicated HTTPS endpoint for one Lambda function, with IAM or no authentication - simple, without throttling or API keys. Lesson 39.12
- FunctionError ("FunctionError": "Unhandled", Sandbox.Timedout)
- The field in an invoke response that says the function failed even though the invoke call itself succeeded (exit code 0); the payload holds the error. Lesson 39.5
- handler (index.handler, lambda_handler, Runtime.HandlerNotFound)
- The function Lambda calls for each invocation, named as file.export (index.handler = the handler export of index.mjs). Lesson 39.5
- high availability (HA, highly available, fault tolerant)
- Keeping a service running through expected failures (an instance, an Availability Zone) automatically, with redundancy in one Region. Lesson 39.27
- hit ratio (cache hit ratio, CacheHitRate, cache hit)
- The share of requests a cache answers itself; the number that says whether a cache helps. Lesson 39.26
- HTTP API (HTTP APIs, quick create, $default route)
- API Gateway's newer, cheaper and simpler API type: routes like GET /items/{id}, Lambda and HTTP integrations, JWT authorizers, auto-deploying stages. Lesson 39.12
- idempotent consumer (idempotency key, idempotent handler)
- A consumer for which processing the same message twice has the same effect as once - needed because queues deliver at least once and batches are retried. Lesson 39.6
- An ECR repository setting that refuses to push a tag that already exists, so a tag always names the same image. Lesson 39.15
- Kinesis Data Streams (Amazon Kinesis, Kinesis, aws kinesis)
- A managed stream of records kept in order per partition key on shards, read by several consumers who can replay it within the retention period. Lesson 39.23
- Lambda destination (on-failure destination, on-success destination, put-function-event-invoke-config)
- Where Lambda sends the record of an asynchronous invocation (event, result or error, attempts): SQS, SNS, Lambda, EventBridge or S3, on success or on failure. Lesson 39.6
- Lambda proxy integration (AWS_PROXY, proxy integration, proxy response)
- The integration that passes the whole request to a function and expects {statusCode, headers, body} back, with body a string; anything else is a 502. Lesson 39.12
- lens (Well-Architected) (Serverless Lens, lenses)
- An extra set of review questions for one kind of workload, such as Serverless, Containers or SaaS. Lesson 39.34
- long polling (ReceiveMessageWaitTimeSeconds, --wait-time-seconds)
- A receive that waits up to 20 seconds for a message instead of returning empty at once: fewer empty receives, lower cost. Lesson 39.1
- MaximumConcurrency (scaling-config MaximumConcurrency, ScalingConfig)
- An SQS event source mapping setting (2-1,000): the most batches it runs at once. Excess messages just wait in the queue - no throttling, unlike reserved concurrency. Lesson 39.6
- message group ID (MessageGroupId, message group)
- The FIFO key that orders messages: messages of one group are delivered in order and one at a time; different groups are processed in parallel. Lesson 39.1
- Missing Authentication Token (MissingAuthenticationTokenException)
- The 403 a REST API returns for a path or method that does not exist on the stage, or an unsigned call to an IAM-authorised method. Lesson 39.12
- multi-site active/active (multi-site, active/active, active-active)
- Both Regions serve users all the time at full capacity; losing one costs almost nothing in time. The most expensive DR strategy. Lesson 39.27
- On-Demand Capacity Reservation (Capacity Reservation, ODCR)
- Guaranteed EC2 capacity of a type in one AZ for as long as you keep it, billed at On-Demand rates whether used or not, with no term. Lesson 39.31
- Parquet (Apache Parquet, .parquet, columnar format)
- A compressed, columnar file format: queries read only the columns they need, so far less data is scanned than with CSV or JSON. Lesson 39.23
- partial batch response (ReportBatchItemFailures, batchItemFailures, partial batch failure)
- The function returns the ids of the records that failed; only those return to the queue and the rest of the batch is deleted. Needs the mapping setting and the code. Lesson 39.6
- partition key (PartitionKey, hot shard, hot partition)
- The value whose hash picks a record's shard: one key's records stay in order on one shard; a key with too much traffic makes a hot shard. Lesson 39.23
- pilot light (pilot-light)
- A DR strategy with data replicated continuously and the core (databases) running in the recovery Region, the application tier off; recovery in tens of minutes. Lesson 39.27
- poison message (poison pill, poison messages)
- A message that fails every time it is processed. Without a DLQ (and partial batch responses) it is retried until it expires and drags its batch with it. Lesson 39.6
- provisioned concurrency (put-provisioned-concurrency-config)
- Execution environments initialised in advance on a version or alias, so those invocations have no cold start; billed while configured. Lesson 39.5
- raw message delivery (RawMessageDelivery, SNS envelope)
- A subscription setting that delivers the published message as it is, instead of wrapped in the SNS notification JSON (Type, MessageId, Message ...). Lesson 39.1
- receipt handle (ReceiptHandle, ReceiptHandleIsInvalid)
- The token a receive returns for one message; delete-message needs it (the MessageId does not work). Each receive gives a new one. Lesson 39.1
- redrive policy (RedrivePolicy, maxReceiveCount, start-message-move-task)
- The SQS attribute naming the DLQ and maxReceiveCount: a message received more often than that moves to the DLQ. start-message-move-task sends messages back after a fix. Lesson 39.1
- refactor (migration) (re-architect, refactoring, rearchitect)
- Rebuilding an application cloud-native (managed services, containers, serverless): the most effort and the most long-term benefit. Lesson 39.36
- rehost (lift and shift, rehosting)
- Moving servers to EC2 as they are, without changing the application: the fastest way to migrate many applications. Lesson 39.36
- relocate (relocating, VMware Cloud on AWS, Amazon Elastic VMware Service)
- Moving a whole platform to AWS without changing it, such as VMware virtual machines that keep their VMware tooling. Lesson 39.36
- replatform (lift, tinker and shift, replatforming)
- Moving an application with one change for a clear benefit, without changing its architecture - typically its database to Amazon RDS. Lesson 39.36
- REPORT line (REPORT RequestId, Billed Duration, Max Memory Used)
- The last log line of each Lambda invocation: Duration, Billed Duration, Memory Size, Max Memory Used, Init Duration on a cold start and a Status on failures. Lesson 39.5
- repurchase (drop and shop, repurchasing)
- Replacing an application with a different product, usually SaaS, and moving its data there. Lesson 39.36
- reserved concurrency (put-function-concurrency, ReservedConcurrentExecutions)
- Concurrency set aside for one function and also its cap; 0 throttles every invocation (the off switch). At least 100 stay unreserved for the rest of the account. Lesson 39.5
- Reserved Instances (RI, RIs, Standard RI)
- The older commitment to an instance type for 1 or 3 years at a discount; zonal RIs also reserve capacity in an AZ. Lesson 39.31
- ResourceInitializationError (CannotPullContainerError, stoppedReason)
- Why an ECS task stopped before its container ran: no network path to ECR / Logs / Secrets Manager, the execution role's missing permissions, or an image that does not exist. Lesson 39.15
- REST API (REST APIs, create-rest-api, resources and methods)
- API Gateway's older, richer API type: resources and methods, usage plans and API keys, request validation, caching, WAF and private APIs. Lesson 39.12
- retire and retain (retire, retain)
- The two strategies that shrink a migration: retire switches off what nobody uses; retain keeps an application where it is for now, for a reason. Lesson 39.36
- Retry (Step Functions) (Retry, ErrorEquals, BackoffRate)
- A state's list of errors to retry, with an interval, a maximum number of attempts and a backoff rate - for transient failures. Lesson 39.20
- right-sizing (rightsizing, right-size, Overprovisioned)
- Matching instance (or function, task) sizes to real use - from Compute Optimizer or CloudWatch history - before committing to anything. Lesson 39.31
- SAA-C03 (Solutions Architect - Associate, AWS Certified Solutions Architect)
- The AWS Certified Solutions Architect - Associate exam: 65 questions, 130 minutes, 720 of 1,000 to pass; secure, resilient, high-performing and cost-optimized design. Lesson 39.38
- saga pattern (saga, compensating step, compensation)
- Keeping a multi-step process consistent without a distributed transaction: each step that changes something has an undo step that runs when a later step fails. Lesson 39.20
- Savings Plans (Compute Savings Plan, EC2 Instance Savings Plan, Savings Plan)
- A 1- or 3-year commitment of dollars per hour for lower prices: Compute Savings Plans follow any EC2 family, Region, Fargate and Lambda; EC2 Instance Savings Plans one family in one Region. Lesson 39.31
- scaled score (compensatory scoring, unscored questions)
- AWS reports results from 100 to 1,000; you pass on the total (compensatory), and 15 of the 65 questions are unscored trial questions you cannot identify. Lesson 39.38
- A Kinesis stream's unit of capacity and order: 1 MB/s or 1,000 records/s in, 2 MB/s out. A shard iterator is a read position in one shard. Lesson 39.23
- AWS secures the cloud itself (hardware, facilities, managed-service software); you secure what you put in it (data, identities, access, configuration) - more for EC2, less for Lambda. Lesson 39.29
- simulate-principal-policy (IAM policy simulator, EvalDecision, implicitDeny)
- Asks IAM whether a user or role would be allowed an action on a resource (allowed, implicitDeny, explicitDeny) without calling the action. Lesson 39.29
- single point of failure (SPOF, single points of failure)
- One component whose failure takes the whole system down - one instance, one AZ, one queue without a DLQ. Lesson 39.27
- six pillars (Well-Architected pillars, operational excellence, performance efficiency)
- The Well-Architected Framework's pillars: operational excellence, security, reliability, performance efficiency, cost optimization and sustainability. Lesson 39.34
- SOA-C03 (CloudOps Engineer - Associate, AWS Certified CloudOps Engineer, SysOps Administrator)
- The AWS Certified CloudOps Engineer - Associate exam (formerly SysOps Administrator): 65 questions, 130 minutes, 720 to pass, no labs; operating systems on AWS. Lesson 39.42
- Spot Instances (Spot, Spot Instance, Spot interruption)
- Spare EC2 capacity at up to 90% off that AWS can reclaim with a two-minute warning - for interruptible, fault-tolerant work. Lesson 39.31
- SSE-KMS (SSE-S3, DSSE-KMS, server-side encryption)
- Encryption at rest by the service: with keys the service owns (SSE-S3, SSE-SQS) or with a KMS key whose policy controls and logs every use (SSE-KMS). Lesson 39.29
- stage (API Gateway) (API stage, create-deployment, deployment (API Gateway))
- A named, deployed snapshot of an API (prod, dev) with its own URL, throttling and logging. Changes take effect only after a new deployment to the stage. Lesson 39.12
- stage throttling (rateLimit, burstLimit, token bucket)
- Rate and burst limits on an API stage, method or route: a token bucket that answers 429 Too Many Requests when empty. Lesson 39.12
- static stability (statically stable)
- A design that keeps working through a failure without having to make changes during it - capacity pre-provisioned, failover by data-plane mechanisms. Lesson 39.27
- task definition (register-task-definition, task definitions, revision (ECS))
- The versioned blueprint of an ECS task: images, CPU and memory, ports, logging, environment and secrets, the execution and task roles. Lesson 39.15
- task execution role (ecsTaskExecutionRole, AmazonECSTaskExecutionRolePolicy, executionRoleArn)
- The role ECS uses before and around a task: pulling the image from ECR, writing the awslogs stream, fetching secrets. Without it the task never starts. Lesson 39.15
- task role (taskRoleArn, ECS task role)
- The role the application inside an ECS task gets for its own AWS calls (S3, SQS ...). A missing permission here is AccessDenied in the app, not a failed start. Lesson 39.15
- task token (.waitForTaskToken, SendTaskSuccess, callback pattern)
- A Step Functions integration pattern that pauses a workflow until something calls back with the token - for human approval or external systems. Lesson 39.20
- TooManyRequestsException (Rate Exceeded, Throttles (Lambda), throttled invocation)
- The error of a throttled synchronous Lambda invocation (HTTP 429); asynchronous events are retried and SQS batches stay in the queue instead. The Throttles metric counts them. Lesson 39.5
- total cost of ownership (TCO, Migration Evaluator, business case)
- Everything a workload really costs - hardware, power, space, licences, people - compared before and after a migration; Migration Evaluator builds the case from measured usage. Lesson 39.36
- usage plan (usage plans, API key, x-api-key)
- A REST API setting that gives each client (identified by an API key) its own rate, burst and quota. API keys meter clients; they do not authenticate them. Lesson 39.12
- visibility timeout (VisibilityTimeout, in flight, in-flight message)
- How long a received SQS message stays hidden from other consumers. If it is not deleted in that time, it becomes visible and is received again. Lesson 39.1
- warm standby (warm-standby)
- A DR strategy with a scaled-down but complete copy running in the recovery Region, scaled up and switched to in minutes. Lesson 39.27
- Well-Architected review (Well-Architected Tool, high risk issue, HRI)
- A structured review of one workload against the six pillars with its team: findings rated as high or medium risk, a ranked improvement plan, and milestones to compare later. Lesson 39.34