Four of the most common CLF-C02 question stems are variations of the same sentence: “A company wants to decouple two components”, “A company needs to notify administrators when…”, “A company wants to react when an event happens in an AWS service”, and “A company needs to coordinate several Lambda functions in order”. Each maps to exactly one of four services, and the exam guide groups them under one heading: Application Integration. Candidates who learn the four services as one-line definitions tend to get the first stem right and guess the rest, because the one-liners overlap (“messaging”, “events”, “serverless”) and the distinctions live one level deeper.
This guide covers that deeper level at the altitude the Cloud Practitioner exam actually uses. Our core services overview gives these services a bullet each, and the business applications guide covers the rest of Task Statement 3.8 (Connect, SES, WorkSpaces, Amplify, AppSync, IoT Core). Here we go through Amazon SQS, Amazon SNS, Amazon EventBridge and AWS Step Functions properly: how each one moves information, what the exam’s favourite comparisons are, and the distractors that sit next to them in the answer options. Service details follow the current AWS documentation as of November 2026.
What the Exam Guide Actually Says
Task Statement 3.8 of the CLF-C02 exam guide is “Identify services from other in-scope AWS service categories”. Its first knowledge bullet reads: “Application integration services of Amazon EventBridge, Amazon Simple Notification Service (Amazon SNS), and Amazon Simple Queue Service (Amazon SQS)”. The matching skill is “Choosing the appropriate service to deliver messages and to send alerts and notifications”.
Appendix A’s in-scope list then names four services under Application Integration:
| In-scope service | What it is in one line | The kind of thing it moves |
|---|---|---|
| Amazon EventBridge | Serverless event bus | Events (something happened) |
| Amazon SNS | Pub/sub messaging and notifications | Messages pushed to many subscribers |
| Amazon SQS | Managed message queue | Messages pulled by one consumer |
| AWS Step Functions | Serverless workflow orchestration | Control flow between steps |
Two observations about that list save marks. First, Step Functions is in scope even though Task Statement 3.8 does not name it; it appears in answer options as “the service that orchestrates”, so you need to recognise it. Second, several services people associate with messaging are not listed under Application Integration: Amazon MQ and Amazon AppFlow are absent from the in-scope list entirely, Amazon Kinesis is listed under Analytics, Amazon API Gateway under Networking and Content Delivery, AWS AppSync under Frontend Web and Mobile, and Amazon SES under Business Applications. They still turn up, but as distractors or as the answer to a differently worded question, and the last section of this guide shows how to tell them apart.
The question style is recognition: scenario in, service name out. You will not configure a queue. You will be asked which service to pick, and occasionally which of two options within a service (Standard vs FIFO, Standard vs Express).
Why Decoupling Is on a Fundamentals Exam
Application integration belongs on the Cloud Practitioner exam because of Domain 1. Task Statement 1.2, “Identify design principles of the AWS Cloud”, lists loose coupling among the principles, and the Well-Architected Framework’s reliability pillar tells you to “decouple components” so that a failure or a slowdown in one does not cascade into the others. The cloud concepts guide covers the framework; this is what the principle looks like in practice.
Consider a web tier that writes orders directly to a processing tier. If processing slows down, web requests back up and users see errors. If processing crashes, orders are lost. Put a queue between them and three things change:
- the web tier returns as soon as the message is accepted, regardless of how busy processing is (asynchronous);
- the processing tier can scale independently, draining the queue at its own pace, and can be replaced or restarted without the web tier noticing (independent scaling and failure isolation);
- a burst of ten thousand orders becomes a backlog that is worked through, not a wave that knocks something over (buffering).
The exam phrases this as “decouple”, “loosely coupled”, “asynchronous”, “buffer”, “absorb spikes” and “components should fail independently”. Every one of those cues points at a queue, a topic or an event bus rather than at a bigger EC2 instance.
Amazon SQS: The Queue
Amazon Simple Queue Service is a fully managed message queue. A producer sends a message to a queue; the message is stored redundantly across multiple Availability Zones; a consumer polls the queue, receives the message, does the work and then deletes it. Nothing is pushed to the consumer. If that sentence is all you remember, you will still get most SQS questions right, because “consumers pull” is the property that separates SQS from SNS.
The facts the exam draws on
| Property | Value | Why it matters in a question |
|---|---|---|
| Delivery model | Consumers poll the queue | SQS never pushes to email, SMS or HTTP; that is SNS |
| Maximum message size | 1 MiB (raised from 256 KB in 2025); larger payloads via the Extended Client Library, which stores the body in Amazon S3 | ”Large payload” questions point at S3 plus a pointer in the message |
| Message retention | 4 days by default, configurable from 1 minute to 14 days | Messages are not kept forever; a backlog older than retention is lost |
| Visibility timeout | 30 seconds by default, up to 12 hours | While a consumer processes a message it is hidden from others; if the consumer fails to delete it in time, it reappears |
| Delay queues | 0 to 15 minutes before a new message becomes visible | ”Delay processing of new messages” |
| Polling | Short polling returns immediately (possibly empty); long polling waits up to 20 seconds for a message | Long polling reduces empty responses and cost; the exam’s “reduce the cost of polling” answer |
| Dead-letter queue (DLQ) | After a message has been received more than maxReceiveCount times without being deleted, it is moved to a DLQ | ”Isolate messages that repeatedly fail” |
| Encryption | Server-side encryption at rest by default (SQS-managed keys, or AWS KMS keys) | Fits the shared responsibility model: AWS encrypts, you choose the key |
| Scale | Practically unlimited throughput for Standard queues; no capacity to provision | ”No servers to manage, scales automatically” |
Standard vs FIFO queues
This is the comparison the exam makes inside SQS, and it is worth a table of its own:
| Standard queue | FIFO queue | |
|---|---|---|
| Ordering | Best-effort ordering; messages may arrive out of order | Strict first-in-first-out order within a message group |
| Delivery | At-least-once; a message may occasionally be delivered more than once | Exactly-once processing; duplicates are removed within a 5-minute deduplication window |
| Throughput | Nearly unlimited | 300 messages per second without batching, 3,000 with batching (higher with high-throughput mode) |
| Name | Any | Must end in .fifo |
| Cue words | ”maximum throughput”, “order does not matter”, “idempotent consumers" | "order must be preserved”, “no duplicates”, “exactly once”, “financial transactions”, “command sequences” |
The trap is to assume FIFO is always better. FIFO costs throughput and a little more money; a question that stresses scale and tolerates reordering wants a Standard queue. A question that says duplicate processing would double-charge a customer wants FIFO.
The patterns worth recognising
- Work queue: producers send jobs, a fleet of EC2 instances or Lambda functions consumes them; an Auto Scaling policy on the queue depth (
ApproximateNumberOfMessagesVisible) adds workers when the backlog grows. This is the canonical “decouple and scale independently” architecture. - Lambda event source mapping: Lambda polls the queue for you and invokes your function with batches of messages, so “process SQS messages without managing servers” is SQS plus Lambda.
- Buffer in front of a database: write requests land in the queue and a consumer applies them at a rate the database can handle.
Amazon SNS: The Topic
Amazon Simple Notification Service is a managed publish/subscribe service. A publisher sends a message to a topic; SNS immediately pushes a copy to every subscription on that topic. There is no polling and, by design, no message store to read from later: SNS retries delivery for a while and can route failures to a dead-letter queue, but a topic is not a queue.
Subscription protocols
The protocol list is itself exam content, because “which service can send an SMS” has one right answer:
| Protocol | Typical use |
|---|---|
| Email / Email-JSON | Alert an operations mailbox |
| SMS | Text an on-call engineer |
| Mobile push | Notifications to iOS, Android and other mobile platforms through their push services |
| HTTP / HTTPS | Call a webhook |
| Amazon SQS | Fan out to one or more queues for durable, asynchronous processing |
| AWS Lambda | Run code in response to the message |
| Amazon Data Firehose | Stream messages into S3, Redshift or OpenSearch |
Two features refine the pattern. Message filtering lets each subscription declare a filter policy so that it receives only the messages whose attributes (or body) match, so one orders topic can feed a high-value queue and a fraud-review Lambda function without the publisher knowing either exists. FIFO topics preserve order and deduplicate, and pair with FIFO queues when ordering must survive the fan-out. Messages are limited to 256 KB.
The fan-out pattern
SNS and SQS are not competitors; the most-tested architecture uses both. A single event, say “order placed”, is published once to an SNS topic. The topic has several SQS queue subscriptions: one for the fulfilment service, one for analytics, one for the email service. Each queue receives its own copy, each consumer works through its queue at its own pace, and adding a fourth consumer means adding a fourth subscription, with no change to the publisher. The exam describes this as “send the same message to multiple services in parallel” or “multiple consumers each need every message”, and the answer is SNS fanning out to SQS.
Contrast the two directly:
| Amazon SQS | Amazon SNS | |
|---|---|---|
| Model | Queue: one message, consumed by one consumer (then deleted) | Topic: one message, copied to every subscriber |
| Direction | Consumers pull | SNS pushes |
| Persistence | Stored until consumed or retention expires (up to 14 days) | Not stored for later retrieval |
| Human-facing delivery | No | Yes: email, SMS, mobile push |
| Cue words | ”decouple”, “buffer”, “process later”, “exactly once”, “work items" | "notify”, “alert”, “subscribe”, “fan out”, “send to multiple” |
Alerts and notifications: the other half of the skill statement
The skill in Task Statement 3.8 is about delivering messages and sending alerts and notifications, and SNS is the glue for the second half. An Amazon CloudWatch alarm’s action is an SNS topic; AWS Budgets notifications go to SNS (or directly to email); AWS Config, GuardDuty findings via EventBridge, CodePipeline approvals and Auto Scaling lifecycle events all end in an SNS topic that an administrator subscribes to. When a question says “notify the operations team when CPU exceeds 80%”, the full answer is a CloudWatch alarm whose action publishes to an SNS topic with an email or SMS subscription. If only one service name is required, it is SNS.
The distractor beside it is Amazon SES, the email-sending service for transactional and marketing mail (receipts, password resets, newsletters) with templates, bounce handling and deliverability tooling. SNS can put a plain-text alert in an inbox; SES sends real email to customers. “Marketing campaign” or “transactional emails to customers” is SES; “alert the administrators” is SNS.
Amazon EventBridge: The Event Bus
Amazon EventBridge is a serverless event bus that receives events and routes them to targets according to rules. It began life as CloudWatch Events, and the older name still appears in some console corners and in older study material; on the exam, EventBridge is the name.
The vocabulary:
- An event is a JSON document describing a change: an EC2 instance changed state, a Step Functions execution failed, a file landed in an S3 bucket, a support case was updated. Events carry a
source, adetail-typeand adetailbody. Events are limited to 256 KB. - An event bus receives events. Every account has a default event bus that automatically receives events from AWS services. You can create custom event buses for your own applications and partner event buses that receive events from integrated SaaS providers.
- A rule matches events with an event pattern (for example,
sourceisaws.ec2anddetail-typeis “EC2 Instance State-change Notification” with stateterminated) and sends matches to up to five targets: Lambda functions, SQS queues, SNS topics, Step Functions state machines, ECS tasks, Kinesis streams, API destinations (HTTP endpoints) and many more. A rule can also transform the event before delivery.
Capabilities that appear as cues:
| Feature | What it adds | Cue |
|---|---|---|
| Scheduled rules / EventBridge Scheduler | Run a target on a cron or rate schedule (Scheduler is the newer, dedicated service with time zones, one-time schedules and flexible windows) | “Run a Lambda function every night at 02:00” |
| SaaS partner sources | Receive events from integrated SaaS applications without polling their APIs | ”React to events from a third-party SaaS tool” |
| Archive and replay | Store events on a bus and replay them later | ”Reprocess last week’s events after a bug fix” |
| Schema registry | Discover and store event schemas and generate code bindings | ”Developers need the structure of events” |
| EventBridge Pipes | Point-to-point connection from one source (SQS, Kinesis, DynamoDB Streams and others) to one target, with filtering and enrichment | ”Connect a queue to a target with filtering and no custom code” |
EventBridge vs SNS
Both can push one message to several places, which is why they share answer options. The distinction the exam expects: EventBridge is about events from AWS services and SaaS applications, routed by content-based rules on a bus; SNS is about publishing messages to subscribers, including people. If the scenario starts with “when an AWS service does something” or “events from a SaaS provider”, it is EventBridge. If it starts with “send a notification”, “subscribers”, “email or SMS”, it is SNS. If it is “fan out one application message to several queues”, it is SNS, though EventBridge could do it too; choose SNS when the question’s language is about subscriptions and notifications.
AWS Step Functions: The Workflow
AWS Step Functions is a serverless orchestration service that coordinates multiple services into a workflow defined as a state machine. Each state does something: run a task (invoke a Lambda function, start an ECS task, call one of thousands of AWS API actions directly), make a choice, run branches in parallel, iterate over items with a Map state, wait for a timer or a callback, or succeed or fail. The definition is written in the Amazon States Language (JSON), or drawn in Workflow Studio, the visual editor, and every execution’s path is visible step by step in the console. Retries, catch blocks and timeouts are built into the definition, so error handling lives in the workflow instead of in every function.
The in-service comparison:
| Standard workflow | Express workflow | |
|---|---|---|
| Maximum duration | Up to one year | Up to five minutes |
| Execution semantics | Exactly-once | At-least-once (synchronous or asynchronous) |
| Pricing | Per state transition | Per execution, duration and memory |
| Fits | Long-running, auditable business processes with human approval steps | High-volume, short event-processing and IoT ingestion |
The cue words are “orchestrate”, “coordinate multiple Lambda functions”, “multi-step process”, “visual workflow”, “human approval step”, “retry logic across steps” and “long-running process”. A scenario that merely needs several functions to run in sequence is better served by Step Functions than by chaining functions, and a question that offers both will reward the orchestration answer.
Step Functions is the one service in the group that is about control flow rather than data movement. A queue moves a message, a topic broadcasts it, a bus routes it; a state machine decides what happens next.
Putting the Four Side by Side
| Amazon SQS | Amazon SNS | Amazon EventBridge | AWS Step Functions | |
|---|---|---|---|---|
| Primitive | Queue | Topic | Event bus and rules | State machine |
| Pattern | Point-to-point, pull | Publish/subscribe, push | Event routing, push | Orchestration |
| Who receives | One consumer per message | Every subscriber | Every matching rule’s targets | The next state |
| Stores messages | Yes (up to 14 days) | No | Only with an archive | Execution history |
| Reaches humans | No | Email, SMS, push | No (via SNS as a target) | No (via tasks) |
| Ordering option | FIFO queues | FIFO topics | No ordering guarantee | Inherent (sequence of states) |
| Typical pair | Lambda or EC2 consumers | SQS queues (fan-out) | Lambda, Step Functions, SQS, SNS targets | Lambda, ECS, SDK integrations |
| One-line cue | ”decouple and buffer" | "notify and fan out" | "react to events" | "coordinate steps” |
The Distractors Next Door
The exam rarely asks “SQS or SNS” with only those two options. It adds services from other categories that sound similar. Learn the boundary for each:
| Service (category) | What it is | How it differs from the application integration four |
|---|---|---|
| Amazon Kinesis (Analytics) | Real-time streaming of high-volume data records (clickstreams, logs, IoT telemetry), retained for replay by multiple consumers | Stream vs queue: “real-time analytics”, “thousands of records per second”, “replay” is Kinesis; “work items processed once” is SQS. The analytics guide drills this trap |
| Amazon API Gateway (Networking and Content Delivery) | Managed front door for REST, HTTP and WebSocket APIs | Synchronous request/response from clients; it can put messages onto SQS or trigger Lambda, but “expose an API” is API Gateway, “decouple” is SQS |
| AWS AppSync (Frontend Web and Mobile) | Managed GraphQL APIs with real-time subscriptions | ”GraphQL” is the giveaway |
| Amazon SES (Business Applications) | Email sending at scale, transactional and marketing | Customer-facing email vs administrator alerts (SNS) |
| Amazon MQ (not in the CLF-C02 in-scope list) | Managed Apache ActiveMQ and RabbitMQ brokers | For migrating applications that already speak JMS, AMQP, MQTT or STOMP without rewriting; new cloud-native applications use SQS and SNS |
| AWS Lambda (Serverless) | Runs code in response to events | The consumer or target, not the integration layer; “run code when a message arrives” is Lambda triggered by SQS, SNS or EventBridge. See the compute guide |
| Amazon CloudWatch (Management and Governance) | Metrics, logs and alarms | The alarm detects; SNS notifies. A question that asks which service sends the message wants SNS |
| Amazon AppFlow (not in the in-scope list) | No-code data transfer between SaaS applications and AWS services | Bulk data flows, not events or messages |
Exam Cue to Service
| The question says | The answer is |
|---|---|
| ”Decouple”, “loosely coupled”, “buffer requests”, “process asynchronously” | Amazon SQS |
| ”Messages must be processed in order, exactly once” | SQS FIFO queue |
| ”Maximum throughput, occasional duplicates acceptable” | SQS Standard queue |
| ”Messages that fail repeatedly should be set aside” | SQS dead-letter queue |
| ”Reduce empty responses and polling cost” | SQS long polling |
| ”Hide a message while a consumer works on it” | SQS visibility timeout |
| ”Send a notification”, “email or SMS to administrators”, “subscribers” | Amazon SNS |
| ”Deliver one message to several queues or services at once” | SNS fan-out to SQS |
| ”Only some subscribers should receive certain messages” | SNS message filtering |
| ”CloudWatch alarm should notify the team” | CloudWatch alarm with an SNS topic action |
| ”React when an AWS service emits an event”, “events from a SaaS application” | Amazon EventBridge |
| ”Run a job on a schedule without servers” | EventBridge Scheduler (scheduled rule) |
| “Replay last week’s events” | EventBridge archive and replay |
| ”Coordinate multiple Lambda functions”, “visual workflow”, “human approval step” | AWS Step Functions |
| ”Long-running process up to a year” / “high-volume, under five minutes” | Step Functions Standard / Express |
| ”Real-time streaming data”, “clickstream” | Amazon Kinesis (not SQS) |
| “Transactional or marketing email to customers” | Amazon SES (not SNS) |
| “Migrate an existing ActiveMQ or RabbitMQ application” | Amazon MQ (not SQS) |
Scenario Drill
| # | Scenario | Answer | Why |
|---|---|---|---|
| 1 | A web tier must keep accepting orders even when the order-processing tier is slow. | Amazon SQS between the tiers | Buffer and decouple |
| 2 | Order-processing must handle each payment exactly once and in order. | SQS FIFO queue | Exactly-once, strict ordering |
| 3 | Image-resizing jobs can run in any order and volume is very high. | SQS Standard queue | Throughput over ordering |
| 4 | Some jobs crash the worker every time and block the queue. | Dead-letter queue with maxReceiveCount | Isolate poison messages |
| 5 | Workers receive many empty responses and the polling bill is rising. | Enable long polling | Wait up to 20 s for messages |
| 6 | A worker takes 90 seconds per message and other workers keep picking up the same message. | Increase the visibility timeout beyond 30 s | Default is 30 seconds |
| 7 | The security team must get an SMS when a GuardDuty finding is high severity. | EventBridge rule on the finding with an SNS topic target (SMS subscription) | Event detected by EventBridge, human notified by SNS |
| 8 | An “order placed” message must reach fulfilment, analytics and email services, each independently. | SNS topic with three SQS queue subscriptions | Fan-out |
| 9 | Only orders above $1,000 should reach the fraud-review queue. | SNS message filtering on the subscription | Filter policy |
| 10 | Administrators must be emailed when EC2 CPU exceeds 80% for 5 minutes. | CloudWatch alarm → SNS topic → email subscription | Alarm detects, SNS delivers |
| 11 | Send receipts and promotional emails to customers. | Amazon SES | Customer email, not alerts |
| 12 | Run a cleanup Lambda function every day at 02:00 UTC. | EventBridge Scheduler | Serverless cron |
| 13 | Trigger a workflow whenever a support case is updated or a SaaS CRM record changes. | Amazon EventBridge (default bus and partner event source) | Events from AWS and SaaS |
| 14 | A bug corrupted a week of processed events; after the fix they must be reprocessed. | EventBridge archive and replay | Replay stored events |
| 15 | A loan application passes through validation, credit check, manual approval and funding, over several days. | Step Functions Standard workflow | Long-running orchestration with human step |
| 16 | Thousands of IoT readings per second each need a three-step transformation lasting milliseconds. | Step Functions Express workflow | Short, high-volume |
| 17 | Five Lambda functions must run in a fixed order with retries between them. | AWS Step Functions | Orchestration, not function chaining |
| 18 | Analyse clickstream data in real time with multiple consumers reading the same records. | Amazon Kinesis | Streaming, not queuing |
| 19 | Expose a REST endpoint that mobile apps call synchronously. | Amazon API Gateway | API, not messaging |
| 20 | Move an on-premises RabbitMQ application to AWS without changing its code. | Amazon MQ | Protocol-compatible broker |
| 21 | A message is 5 MB. | Store the payload in S3 and send a pointer (Extended Client Library) | SQS limit is 1 MiB |
| 22 | A message must remain available for consumers for up to two weeks. | SQS with retention set to 14 days | Maximum retention |
| 23 | Which service in the scenario “alarm → notification → on-call phone” actually sends the SMS? | Amazon SNS | Delivery is SNS, detection is CloudWatch |
| 24 | Developers need to know the exact JSON structure of EC2 state-change events. | EventBridge schema registry | Schema discovery |
Mistakes That Cost Marks
- Picking SNS for “decouple”. Decoupling and buffering are queue words; SNS stores nothing.
- Picking SQS for “notify”. SQS cannot email, text or push to a phone.
- Confusing SNS with SES. Alerts to administrators are SNS; email to customers is SES.
- Confusing SQS with Kinesis. Queues hand each message to one consumer and delete it; streams keep ordered records for multiple consumers and real-time analytics.
- Assuming FIFO is always right. It trades throughput for ordering; read the scenario’s priorities.
- Forgetting that Step Functions is in scope. It is absent from the task statement’s text but present in the in-scope list and in answer options.
- Treating CloudWatch as the notifier. CloudWatch alarms trigger; SNS delivers.
- Treating EventBridge and SNS as interchangeable. Events from AWS services and SaaS go through EventBridge; subscriber notifications go through SNS; the two are often chained.
Practice Recognising the Cue
These services are learned by exposure to scenarios, not by rereading definitions. The domains breakdown shows where Task Statement 3.8 sits, the study plan schedules it, and the free practice questions and free CLF-C02 sample exam show how the stems are worded. When you want full-length timed practice, Sailor.sh’s AWS Certified Cloud Practitioner (CLF-C02) Mock Exam Bundle includes eight 65-question exams whose Domain 3 items pair SQS, SNS, EventBridge and Step Functions against Kinesis, SES, API Gateway and Amazon MQ in the same answer sets, with an explanation of why each distractor is wrong. By the third exam the cue words become automatic, which is the whole point.
Frequently Asked Questions
What is the simplest way to remember SQS vs SNS?
A queue holds messages until one consumer pulls and deletes them; a topic pushes each message to every subscriber immediately. “Buffer” is SQS, “broadcast” is SNS.
Does the Cloud Practitioner exam ask about FIFO queues?
Yes, at recognition level: FIFO preserves order and processes exactly once, Standard offers higher throughput with at-least-once delivery and best-effort ordering. You will not be asked about message group IDs or deduplication IDs.
Is EventBridge the same as CloudWatch Events?
EventBridge is the successor and superset of CloudWatch Events, built on the same service. Current exam questions use the EventBridge name.
When is Step Functions the answer instead of chaining Lambda functions?
Whenever the scenario mentions coordinating several steps, retries across steps, branching, parallel work, waiting for approval, or a visual view of the process. Chaining functions directly hides the workflow and spreads error handling across every function.
Where does Amazon MQ fit if it is not in scope?
It appears as a distractor. It is a managed broker for applications that already use ActiveMQ or RabbitMQ protocols. For new cloud-native designs the exam expects SQS and SNS.
How large can an SQS message be?
Up to 1 MiB since 2025 (previously 256 KB). SNS messages and EventBridge events are limited to 256 KB. Larger payloads go to Amazon S3 with a reference in the message.
Can a single AWS service trigger both an SNS notification and a Lambda function?
Yes. An EventBridge rule can have up to five targets, so one event can notify a topic and invoke a function. SNS topics can also fan out to Lambda functions and SQS queues at the same time.
Which of these services is the right answer for “send alerts to administrators”?
Amazon SNS, usually fed by a CloudWatch alarm or an EventBridge rule. Email and SMS subscriptions are what make it the human-facing service in the group.
Conclusion
Application integration on the CLF-C02 is four services with four verbs: SQS buffers, SNS notifies, EventBridge routes, Step Functions orchestrates. Learn the properties that make each one distinct (pull vs push, stored vs not stored, events vs messages, data vs control flow), the two internal choices (Standard vs FIFO queues, Standard vs Express workflows), and the five neighbours that are not in the group (Kinesis, SES, API Gateway, Amazon MQ, CloudWatch), and the scenarios resolve themselves. Decoupling is the principle the exam is really testing; these are simply the services that implement it.