AWS Certified Cloud Practitioner (CLF-C02) #6 Domain 3-1 Core Services — Compute and Storage

With #5 we finished Domain 2. Now into the domain that carries the heaviest weight on the exam — Cloud Technology and Services (34%).

Almost every question in this domain takes the shape “Which service best fits this scenario?”. Rather than memorizing every feature of every service, it’s far more efficient to classify the kind of workload each service solves.

This post covers the front half of Domain 3 — compute and storage.

Compute services at a glance #

EC2 — Elastic Compute Cloud #

The most basic service providing virtual machines. We covered it in Basics Track #2.

Instance families #

EC2 instance names carry a family + generation + size (m5.large, c6i.xlarge).

You don’t need to memorize every family for the exam. “GPU workloads go to G/P”, “memory-heavy goes to R”, “general-purpose is M” is enough.

EC2 instance pricing models (briefly) #

We cover pricing in detail in #8. The essentials:

Lambda — serverless functions #

A service that runs your code with no server management.

Exam patterns #

• “Automatically generate a thumbnail whenever an image is uploaded to S3” → Lambda + S3 events
• “Run a REST API with no server management” → Lambda + API Gateway
• “Data processing that takes more than an hour” → Lambda doesn’t fit (15-minute limit). Use ECS/Batch
• “Highly irregular usage, only fires occasionally” → Lambda (charged only for what you use)

ECS, EKS, Fargate — containers #

ECS’s two launch modes #

Exam patterns #

• “Run Docker containers without managing nodes” → Fargate
• “Use the standard Kubernetes interface and have AWS manage it” → EKS
• “Just run containers on AWS, nothing fancy” → ECS
• “We use Kubernetes on-prem and want the same tooling on AWS” → EKS (portability)

Elastic Beanstalk #

A service that deploys apps as a PaaS. Upload your code as a zip and AWS provisions EC2, ELB, and Auto Scaling automatically.

Supported platforms: Node.js, Python, Java, Ruby, Go, .NET, PHP, Docker, and more.

Exam patterns #

• “Deploy a Node.js app as fast as possible with minimal infrastructure setup” → Elastic Beanstalk
• “Keep existing EC2/ELB/Auto Scaling and just automate deployment” → CodeDeploy fits better than Beanstalk

Lightsail #

Provides a simple virtual server plus prepackaged apps (WordPress, LAMP, MEAN) on flat-rate pricing.

EC2 vs Lightsail #

• EC2 — Powerful and flexible, but requires infrastructure knowledge. Usage-based pricing
• Lightsail — Simple. Flat rate. For beginners and small workloads

AWS Batch #

A managed service for batch jobs. Automatically schedules and retries hundreds to tens of thousands of jobs.

Exam patterns #

• “Batch-encode hundreds of thousands of videos” → AWS Batch
• “Process massive data overnight” → Batch
• “A short, one-shot data transformation” → Lambda

Outposts, Local Zones, Wavelength #

For the exam, just map use cases:

• Data must not leave the corporate data center → Outposts
• Gaming/media needing single-digit ms latency → Local Zones
• Minimum latency for 5G mobile users → Wavelength

Compute workload mapping #

Storage services at a glance #

S3 (Simple Storage Service) #

The standard for object storage. Files are accessed by URL.

Characteristics #

• Infinite scale — Practically no limit on object count or total capacity
• Up to 5 TB per object
• 99.999999999% (11 9s) durability — Automatically replicated across multiple facilities
• Bucket names are globally unique
• Looks global, but data lives in a region

S3 Storage Classes #

The most frequently tested area. Tiers based on access frequency.

Exam patterns #

• “Accessed about once a month but must be retrieved immediately” → S3 Standard-IA
• “Access pattern is unpredictable, optimize automatically” → Intelligent-Tiering
• “7-year legal retention, almost never accessed, lowest cost possible” → Glacier Deep Archive
• “Reproducible backup, minimize cost” → One Zone-IA

S3 Lifecycle Policy #

A rule that automatically moves objects to another class after a time threshold.

Object created → after 30 days move to Standard-IA → after 90 days to Glacier → after 1 year delete

S3 security #

Common S3 scenarios on the exam #

• “Static website hosting” → S3 (Static Website Hosting)
• “Upload a single object larger than 5 GB” → Multipart Upload
• “Grant temporary download access to a third party” → Pre-signed URL
• “Replicate S3 objects to another region” → Cross-Region Replication (CRR)
• “Replication within the same region” → Same-Region Replication (SRR)

EBS (Elastic Block Store) #

EC2’s disk.

Exam patterns #

• “EC2’s boot disk” → EBS
• “EC2 in another AZ mounts the same disk” → Not possible. Use EFS
• “EBS backup” → Snapshot (stored in S3)

EFS (Elastic File System) #

File storage that multiple EC2 instances mount simultaneously. Linux only (NFS).

Exam patterns #

• “Multiple EC2 instances share files with read/write” → EFS
• “Shared files for Windows servers” → EFS doesn’t fit. FSx for Windows

FSx #

Managed file systems for specialized workloads.

For the exam, just map:

• Windows SMB share → FSx for Windows
• HPC, ML → FSx for Lustre

Storage Gateway — hybrid storage #

A gateway that bridges on-prem and AWS storage.

Exam scenario: “Move on-prem backup tape systems to the cloud” → Tape Gateway.

Snow Family — bulk data transfer #

Petabyte-scale data is shipped on physical devices rather than over the internet.

Exam scenario: “Move 500 TB from the corporate data center to AWS in one shot, internet is slow” → Snowball Edge.

AWS Backup — unified backup #

Consolidates backups across services (EBS, RDS, EFS, DynamoDB, FSx, and more) into a single policy.

Storage workload mapping #

Common traps #

1) Picking Lambda as the answer to everything compute #

Lambda has a 15-minute limit. Long jobs go to ECS/Batch.

2) EBS Multi-AZ #

EBS is bound to a single AZ by default. EC2 in another AZ cannot mount the same disk. For sharing, use EFS.

3) EFS and Windows #

EFS is Linux NFS only. For Windows, use FSx for Windows.

4) Assuming all S3 Glacier classes are instantly retrievable #

Glacier classes have different retrieval times. Only Instant Retrieval is immediate; the others take minutes to hours.

5) Trying to memorize every EC2 instance family #

T, M, C, R, G, and P are plenty. You don’t need to memorize every generation and size.

6) Dismissing Lightsail #

In small-scale, flat-rate scenarios, Lightsail is the right answer more often than you’d expect.

Wrap-up #

What this post locked in:

• Compute — EC2 (VM) / Lambda (serverless) / ECS, EKS, Fargate (containers) / Beanstalk (PaaS) / Lightsail (small flat-rate) / Batch (batch jobs) / Outposts (hybrid)
• Workload → service mapping is the core
• EC2 instance families — T, M, C, R, G, P and friends classified by purpose
• Storage — S3 (object) / EBS (block) / EFS (Linux file) / FSx (specialized file) / Storage Gateway (hybrid) / Snow (offline transfer)
• S3 Storage Classes — Standard / Intelligent-Tiering / Standard-IA / One Zone-IA / three Glacier flavors (Instant, Flexible, Deep Archive)
• Lifecycle Policies for automatic tier transitions
• Traps — Lambda’s 15-minute limit, EBS’s AZ binding, EFS being Linux-only, Glacier retrieval times, over-memorizing instance families

Next — Domain 3-2 Networking and Databases #

On to the back half of Domain 3.

#7 Domain 3-2 Core Services — Networking and Databases covers VPC, Subnet, Route 53, CloudFront, the four flavors of ELB, VPN, Direct Connect, and Global Accelerator, plus the classification and use-case mappings for RDS, Aurora, DynamoDB, ElastiCache, and Redshift.