A CLF-C02-style practice question describes a directory of config files that 200 web servers all need to read and update, and two of the options, S3 and EFS, both look like the answer. The exam guide describes its distractors as "generally plausible responses that match the content area" (exam guide). The exam tests whether you can read the question and rule out the plausible wrong service. This page is a service-selection reference for the AWS Certified Cloud Practitioner (CLF-C02) exam. Each row below gives you the phrase in the question that decides the answer, the service that phrase points to, the tempting wrong service, and the one-line mechanism that separates them. The rows are grouped under the exam's four scored domains, so you can weight your review by what each area is worth.
What the CLF-C02 exam measures
CLF-C02 validates what AWS describes as "overall knowledge of the AWS Cloud, independent of a specific job role." Two of the tasks the exam guide lists as measured are exactly what this page trains: "Describe and position the core AWS services, including compute, network, database, and storage services" and "Identify AWS services for common use cases." Per the current exam guide and the certification page:
- 90 minutes, 65 questions. 50 questions affect your score; 15 are unscored and collected for future use. Unanswered questions are scored as incorrect, and there is no penalty for guessing.
- Multiple choice questions have one correct answer and three distractors. Multiple response questions have two or more correct answers out of five or more.
- Your results are reported as a scaled score of 100-1,000. The minimum passing score is 700. Scoring is compensatory, so you do not need to pass each section, only the exam as a whole.
- The exam is 100 USD, taken at a Pearson VUE test center or online proctored. The certification is valid for 3 years, and after you earn one AWS certification, you get a 50% discount on other AWS certification exams.
The four scored domains and their weightings (exam guide):
- Cloud Concepts: 24% of scored content
- Security and Compliance: 30% of scored content
- Cloud Technology and Services: 34% of scored content
- Billing, Pricing, and Support: 12% of scored content
Recheck the exam guide and the certification page before you book. AWS updates both. The certification page currently notes that the Italian and German language versions retire after December 31, 2026. Recheck that line after that date.
How to read each table
Read each row in three passes. First, find the one phrase in the question that decides the answer. "Every instance reads and updates the same files." "Users in Europe need lower latency." Second, figure out the distractor before you read the options. The exam guide defines a distractor as a generally plausible response that matches the content area. In a service question, that plausible option is usually the service whose name matches the topic of the question, so working it out ahead of time takes it off the table. Third, check the mechanism. The mechanism is the one-line reason from the official documentation, and if you can state it, you can defend your pick under time pressure.
Worked case (reading the question). The scenario from the opening of this page: a company runs a web application on 200 EC2 instances and needs one directory of config files that every instance reads and updates. The options are S3, EFS, and EC2 instance store. The cue is "every instance reads and updates the same files". Amazon EFS is a file system that compute instances mount, with massively parallel access from compute, and the service manages the file storage infrastructure for you (EFS docs). S3 stores objects retrieved by key, not a file system you mount (S3 docs). Instance store holds temporary data that is deleted when the instance stops, hibernates, or terminates (EC2 docs), so a directory on instance store would be lost when any of the 200 instances stops or is replaced. What tempts you here is the word storage. S3 carries the storage name, and a question that mentions files at all pulls toward it. The mechanism that breaks the tie is mount versus retrieve. Instances mounting one shared file system point to EFS.
A practice test shows which of these decisions are your gaps: infrastructure placement, compute model, storage type, database model, monitoring versus auditing, or billing. Study and retest those rows before the next set.
Global infrastructure: where the workload sits
Cloud Concepts is 24% of scored content. The decisions in this domain are placement decisions: where the workload runs, and how content reaches the users. Per the AWS global infrastructure page, the AWS Cloud spans 39 geographic Regions and 124 Availability Zones, and each Region consists of at least three independent, physically separated Availability Zones (global infrastructure page).
| The situation | Pick | Common wrong pick | What decides it |
|---|---|---|---|
| Users in another country or continent need lower latency, or the workload must stay isolated from failures in another Region | A separate AWS Region in that geography | Another Availability Zone, or another VPC, in the current Region | An AZ lives inside its Region. A second AZ adds resiliency inside the same geography and does not move the workload (global infrastructure page) |
| Files, media, or web content served close to users around the world | Amazon CloudFront | Deploy the application in every Region | CloudFront caches objects in edge locations, a worldwide network of data centers, and routes each request to the lowest-latency edge location. The origin, an S3 bucket or an HTTP server, keeps the definitive copy, so the workload stays where it was (CloudFront docs) |
The tempting option here is another Availability Zone in the same Region. An AZ is a resiliency unit inside one Region, not a geographic move, so it does not change where the workload runs. If the question adds users in Europe, a European Region is the only option that changes where the workload actually runs. CloudFront speeds delivery from European edge locations, but the origin stays where it was.
Compute: which service runs the code
Cloud Technology and Services is 34% of scored content, the largest of the four domains. The rows below cover the compute model: continuous or long-running work, and short work triggered by events. The same domain's storage and database rows appear in their own sections below, and its network services have no rows in this article.
| The situation | Pick | Common wrong pick | What decides it |
|---|---|---|---|
| Runs continuously, always-on, dedicated capacity, or a long-running process (an app server, a monthly batch job) | Amazon EC2 | Lambda | An EC2 instance is a virtual server you run for as long as you need, and you scale capacity up and down as usage changes (EC2 docs) |
| Triggered by events: an API call, a file lands in S3, a queue message; short bursts of work | AWS Lambda | EC2 | Lambda functions run code in response to events without managed servers, up to 15 minutes per invocation, with pay-per-use billing and no upfront commitments (Lambda docs) |
Any workload that runs code is compute, so a question that says run a process does not select EC2. What selects it is the lifecycle of the work. Continuous or long-running work runs on a server you keep. Short work triggered by events runs on a function AWS starts for you. If the process can run longer than 15 minutes, Lambda is out, because a function runs for at most 15 minutes.
Storage: where the data lives
| The situation | Pick | Common wrong pick | What decides it |
|---|---|---|---|
| Many instances need to read and write the same files at once; a shared file system | EFS | S3 | EFS is a file system instances mount (NFSv4, POSIX permissions), with massively parallel access from compute; S3 stores objects retrieved by key (EFS docs, S3 docs) |
| Objects rather than files: media, backups, a data lake, a static website | Amazon S3 | EFS or an EC2 instance store | S3 is object storage built for websites, mobile applications, backup and restore, data lakes, and archive (S3 docs) |
| Data kept for years, rarely accessed, at the lowest cost | An S3 Glacier storage class (Instant Retrieval, Flexible Retrieval, or Deep Archive), usually via a lifecycle rule | Keep it in S3 Standard | S3 tiers storage by access frequency; the Glacier classes exist to archive at the lowest costs, and lifecycle configurations move objects between classes (S3 docs) |
| A persistent disk attached to an EC2 instance | Amazon EBS | Instance store | The EC2 guide describes EBS volumes as persistent storage volumes for your data; instance store holds temporary data deleted when the instance stops, hibernates, or terminates (EC2 docs) |
A question that says store files pulls toward S3, because storage is in the name. What decides it is how the workload accesses the data: mounted and shared across instances (EFS), retrieved as objects (S3), or attached to a single instance as persistent storage (EBS).
Databases: RDS or DynamoDB
| The situation | Pick | Common wrong pick | What decides it |
|---|---|---|---|
| Relational model, SQL, or joins across tables; an existing application on MySQL, PostgreSQL, Oracle, SQL Server, Db2, or MariaDB | RDS | DynamoDB | RDS runs those relational engines and manages backups, patching, failure detection, and recovery; the docs recommend it as the default for most relational deployments (RDS docs) |
| Key-value or document access patterns, predictable per-item performance, capacity that should scale with traffic | DynamoDB | RDS | DynamoDB is a serverless NoSQL database that supports key-value and document data models, does not support the JOIN operator (denormalize the data instead), and scales on-demand capacity with requests, down to zero (DynamoDB docs) |
Both are databases, so the data model decides. If the question's application performs joins or runs existing SQL, DynamoDB cannot express the query, because it has no JOIN operator, so RDS is the answer. If the question describes per-item lookups at scale, the DynamoDB documentation's own example is a shopping cart that keeps single-digit millisecond performance from 10 users to 100 million, and the relational engine is the heavier tool.
Monitoring and auditing: who did what
Security and Compliance is 30% of scored content, and the exam guide lists "understand and explain the AWS shared responsibility model" among the tasks the exam validates. In a service-selection question, that model shows up as the question of who manages what, and the EFS and DynamoDB pages state their share of it. EFS says the service manages all the file storage infrastructure for you. DynamoDB says it handles setup, configurations, maintenance, high availability, hardware provisioning, security, backups, and monitoring. The service-selection version of that question is the monitoring-versus-auditing pair:
| The situation | Pick | Common wrong pick | What decides it |
|---|---|---|---|
| Track metrics, set an alarm on a threshold, view dashboards, monitor logs | Amazon CloudWatch | CloudTrail | CloudWatch is the monitoring and observability service; it collects and visualizes metrics, logs, and traces and sets alarms (CloudWatch page) |
| An audit or compliance question: who made which API call, in the account, and when | AWS CloudTrail | CloudWatch | CloudTrail records actions taken by a user, role, or AWS service as events for auditing, governance, and compliance; Event history keeps a free 90-day record of management events automatically, and a trail delivers the record to an S3 bucket (CloudTrail docs) |
Both services produce logs, and a question that says record activity does not select either one by itself. What decides it is what the record is for: operational monitoring of a workload (CloudWatch) or an audit of account activity that answers who or what took which action and when (CloudTrail).
Billing, pricing, and support: the domain worth 12%
Domain 4 is 12% of scored content, and the exam guide ties it to the task of understanding "AWS Cloud costs, economics, and billing practices." The verified facts a CLF-C02 candidate should hold:
- The exam is 100 USD, and the certification page links to exam pricing for additional cost information, including foreign exchange rates. The 3-year validity and the 50% discount from the first section complete this list.
- AWS's official practice question set, pretest, and practice exam are part of the Exam Prep Plan on the certification page.
- The service pages above state their billing model in one line: Lambda is pay-per-use with no upfront commitments, and DynamoDB's on-demand mode bills the requests you make.
Billing questions test whether you can read those mechanics. Your next set should include one question on a pay-per-use model, like Lambda's, and one on request-based billing, like DynamoDB on-demand, because a question that says you are billed for usage can point to either.
How to know you have closed a gap
You have closed a gap when you still pick the right service after the wording changes. Practice sets phrase the same decision differently. "Shared file system" and "every instance needs the same directory" are the same EFS cue. As a CramHQ rule of thumb, treat a gap as closed when you pick the right service on a timed set you have not seen and can state the mechanism out loud. If you can only name the answer and not the mechanism, the gap is still open. Retest those weak areas close to where the practice test found them. If the compute row is a weak area, your next set should include both an event-triggered question and a continuous-workload question, because a similar phrase can point to either service.
The free assessment for these decisions
Take the free AWS CLF-C02 assessment on CramHQ. It gives you a Readiness Report and your top gaps. The full CLF-C02 path is also free, with no card required, and includes original practice questions with explanations, flashcards, targeted Pass Plan repairs, 5 practice tests, and the Final Timed Simulation. It lives at the course page. When the Readiness Report shows a gap, open the row above, reread the mechanism line, and retest the nearest questions.