Cloud Economics, Migration Strategies and the Cloud Journey

Domain 1 of the CLF-C02 exam carries 24% of the scored content, the second-largest domain on the paper, and it is the one domain where AWS asks you to argue in the language of a business case rather than a service feature. Everything else in the exam (security, global infrastructure, services, billing) assumes you already accept the premise; Domain 1 is the premise. Candidates who arrive with a long list of service names but no vocabulary for fixed versus variable cost, capex versus opex, or "why rehost and later replatform" lose marks they never expected to be answering for.

In this lesson you will:

• state the six advantages of cloud computing in AWS's own wording;
• separate economies of scale, global reach, speed, elasticity, scalability and high availability;
• walk the six Well-Architected pillars and the six general design principles, and show how the pillars trade off against each other;
• map the six AWS CAF perspectives to owners and to the four examinable outcomes;
• choose between the 6 Rs and the 7 Rs migration strategies with a decision tree, and name the tools behind database replication;
• place the Assess → Mobilize → Migrate & Modernize journey and its supporting resources on a single map;
• convert a fixed-cost, capex on-premises estate into a variable-cost, opex cloud model;
• build a TCO and ROI comparison with worked arithmetic you can reproduce in 90 seconds;
• decide between BYOL and license-included, and order rightsizing before discounts;
• explain how the responsibility line moves across EC2, RDS and Lambda;
• read two real AWS migration case studies and handle their numbers honestly;
• answer 10 exam-style questions plus three interactive checks;
• then extend the story set with Netflix, Box and FarEye, work the 2026 updates box, and drill a case-study matching check plus two extra questions (clf-02-q11, clf-02-q12).

---

1. The value proposition of the AWS Cloud (task 1.1)

1.1 The six advantages, verbatim

AWS publishes six advantages of cloud computing in a whitepaper, and exam questions quote the phrases almost word for word. Learn the bolded trigger phrase for each row; that is the string the question stem will reproduce.

AWS also states it operates over 200 fully featured services (as of Oct 2026): a figure worth remembering because distractors like "over 1,000 services" are easy to eliminate.

1.2 Example E1: economies of scale, arithmetically

Economies of scale are structural, not a promotion. The cloud provider aggregates demand across customers, so it buys power, hardware and floor space at a price no single company can reach, and passes part of the difference into a lower variable rate.

Two customers, each needing a peak of 10 units, but their peaks are anti-correlated (one is busy when the other is quiet):

1.3 Global reach, speed of deployment and agility

• 📚 Did you know? As of Oct 2026 AWS reports 39 geographic Regions and 124 Availability Zones, with each Region required to carry a minimum of three isolated, physically separate AZs, all within 100 km (60 miles) of each other. An Availability Zone is defined as "one or more discrete data centers, each with redundant power, networking, and connectivity", so an AZ is not a rack, and a Region is not an AZ.

1.4 Example E2: "stop guessing capacity" in numbers

A team forecasts a peak of 100 vCPU. On-premises practice is to buy 20% to 50% over the peak-predicted demand so the forecast cannot bite:

That is what "stop guessing capacity" means: instead of buying for the peak and sitting idle, you provision to measured demand in minutes and let elasticity absorb the spike. The same logic repeats at every later layer of this lesson: idle capacity is the enemy in rightsizing, in licensing and in the TCO table.

---

2. Design principles: the AWS Well-Architected Framework (task 1.2)

2.1 What the Framework is

The AWS Well-Architected Framework documents best practices for designing and operating secure, reliable, efficient, cost-effective, and sustainable workloads. The exam-relevant detail about the process is this: a Well-Architected review is "a constructive conversation... not an audit mechanism"; you will never be asked to "pass" one, and the AWS Well-Architected Tool that runs it is free and on the in-scope service list.

2.2 The six pillars

2.3 The six general design principles

All six principles do the same underlying move: they convert upfront, fixed, idle spend into metered, on-demand, evidence-driven spend.

2.4 Example E3: the pillars are not interchangeable

Task 1.2 explicitly asks you to identify differences between the pillars. They conflict, and the exam tests whether you notice:

There is no pillar that wins automatically. The correct exam posture is a trade-off that is deliberate and measured, not a claim that one pillar is "the most important". Note also that Sustainability and Cost Optimization both talk about reduction, and Security and Reliability both talk about resilience: similar words, different pillars, different owners.

• 📚 Did you know? The Cloud Adoption Framework (Section 3), the Well-Architected Framework (this section) and the Cloud Value Framework (Section 6) are three different lists of five or six items that candidates merge constantly: Well-Architected has 6 pillars about workload quality, CAF has 6 perspectives about organizational readiness, and the Cloud Value Framework has 5 pillars about business value: of which TCO is only pillar 1 of 5.

---

3. The AWS Cloud Adoption Framework (task 1.3)

3.1 Six perspectives, six owners

The AWS Cloud Adoption Framework (AWS CAF) groups its capabilities in six perspectives. Each perspective is a conversation with a different part of the organization; that is why the exam asks you to match a perspective to a stakeholder or a concern.

Published capability counts (as of Oct 2026): People 7, Governance 7, Platform 7, Security 9, Operations 9. AWS does not publish a Business-perspective count in the sources verified for this lesson, so do not memorize a total.

3.2 The four examinable outcomes

Task 1.3 names the outcomes you must recognize. A CAF-led program is judged on:

1. Reduced business risk
2. Improved environmental, social, and governance (ESG) performance
3. Increased revenue
4. Increased operational efficiency

3.3 One map: CAF → MAP → Well-Architected

These three frameworks appear together in almost every migration question. They answer three different questions, in this order:

---

4. Migration strategies: the 6 Rs (task 1.3)

4.1 The strategies and their glosses

AWS's older and still exam-popular framing is the 6 Rs, built on the five Rs that Gartner outlined. The current AWS Prescriptive Guidance framing is the 7 Rs, which adds Relocate. Both appear in study material; the exam guide itself never enumerates either list.

4.2 Example E4: sorting a 40-application portfolio

An enterprise assesses 40 applications. Using AWS's own order-of-magnitude guidance that "as much as 10% to 20% of an enterprise IT portfolio is no longer useful and can be turned off":

Sequencing rule for large migrations: rehost, replatform, relocate and retire first; modernize after. Paying a migration factory to move 40 applications in parallel is how programs blow their budget; retiring 4 of them before anyone writes a line of code is how programs fund themselves.

4.3 The two named tools behind the strategies

4.4 Example E5: the Snowball reality check

For petabyte-scale transfer over a constrained link, the classic answer is a physical device: a single Snowball Edge Storage Optimized device moves approximately 210 TB (as of Oct 2026; verify current specs before use), and a job must be completed within 360 days of being prepared.

Two caveats you must hold at the same time:

• 📚 Did you know? Snowball Edge Storage Optimized ships with about 210 TB of usable NVMe (as of Oct 2026) while the Compute Optimized variant trades storage for silicon, roughly 28 TB NVMe against 416 GB of memory. The device is on neither the in-scope nor the out-of-scope service list for CLF-C02 (both lists are explicitly "non-exhaustive"), so whether it is examinable cannot be verified, treat it as knowledge, not as a certainty.

---

5. The cloud migration journey and its resources (task 1.3)

5.1 The three phases

AWS's migration program runs in three phases, and the exam asks for the order:

Read literally: Assess → Mobilize → Migrate & Modernize. Mobilize sits between the business case and the first wave, because the gaps found in Assess have to be closed before migration starts.

5.2 The resources that support the journey

---

6. Cloud economics (task 1.4)

6.1 Fixed cost versus variable cost

Fixed cost exists whether or not the service runs: owned hardware, floor space, salaried staff. Variable cost tracks consumption. The single most-quoted sentence in Domain 1 is: "The cloud trades fixed expense for variable expense; you pay only when you consume computing resources."

Example E6: twelve months, illustrative figures (as of Oct 2026; teaching numbers only, not AWS list prices):

The on-premises row is a flat line because you paid for it in the capital year: traffic has nothing to do with it. The cloud row can drop to zero for a workload you switch off.

6.2 The cost drivers of an on-premises environment

Task 1.4 asks specifically for "costs associated with on-premises environments". Every row below has to appear in a TCO model or the model is wrong:

6.3 Example E7: the hidden on-premises inputs

Most on-premises models fail by counting hardware only. Here are ten servers over five years, with all figures illustrative (as of Oct 2026, teaching numbers, not AWS list prices):

AWS's own guidance on accurate comparisons is blunt: capital expenditures usually have a refresh cycle of five years, equipment is procured six to nine months in advance, and customers overprovision 20% to 50% over peak-predicted demand. A model that omits labour, refresh and facilities understates on-premises cost; which is why the exam treats "compare server purchase price to an hourly rate" as the classic flawed TCO model.

6.4 Capex to opex

6.5 Example E8: TCO and ROI, worked

TCO is the all-in, multi-year comparison of acquisition and operating costs, hardware, software, facilities, labour and capital, on-premises versus AWS. It is not a price list, and AWS explicitly frames it as only pillar 1 of 5 of the Cloud Value Framework.

All dollar figures in the table below are illustrative teaching numbers created for this lesson (as of Oct 2026). They are NOT AWS list prices and must not be quoted as such.

Now the two numbers an interviewer (or an exam stem) actually wants:

Sanity check against a published figure: AWS cites IDC (2022) claiming 50% lower 5-year cost of operations on AWS: a customer/AWS-cited, unaudited claim, and our illustrative −49.9% lands in the same neighbourhood by construction, not by coincidence. Always attribute such figures ("IDC for AWS, 2022"), never present them as guaranteed.

6.6 Licensing: BYOL versus license-included

License-included means Amazon EC2 instances that bundle the licence cost into the compute cost, AWS is responsible for licensing compliance, and you pay only while the instance runs. BYOL means you bring your own licence and media, you own compliance. AWS License Manager is a free tool that tracks both across AWS, on-premises and other clouds.

Example E9: 30% uptime workload, illustrative rates (as of Oct 2026; teaching figures only):

Rule: variable or intermittent uptime → license-included; high, steady uptime plus existing discounted licences → BYOL. Remember that a Windows licence can represent 50% or more of compute cost (AWS, 2022-04-28), so this is not a rounding decision.

6.7 Rightsizing

Rightsizing means matching instance size and type to measured utilization while still meeting performance requirements. It is not "pick the cheapest instance", and the order of operations matters.

Example E10: from 12 vCPU to 4 vCPU, illustrative rate (as of Oct 2026; teaching figure only: $0.06 per vCPU-hour):

AWS Compute Optimizer "provides you with rightsizing recommendations and identifies idle resources." Only after rightsizing do you choose a commitment model; because committing to a discount on an oversized instance simply buys a discount on waste. AWS documents Savings Plans as "up to 72% lower in some cases" versus on-demand and Spot as "up to a 90% discount from on-demand pricing" (both AWS, 2022-04-28; verify current rates before use).

• 📚 Did you know? Published utilization figures are the economic core of this lesson: on-premises servers average 12% to 18% utilization, more than 30% of servers run under 10% utilization, and an on-premises data center is described as "under 15%" (AWS blogs, 2022-04-28 and 2024-10-01). That is why rightsizing first, commitment second is the canonical FinOps order on the exam.

---

7. Automation, managed services and the responsibility shift (task 1.4)

7.1 Automation benefits

Infrastructure as code (IaC) means provisioning and supporting computing infrastructure using code instead of manual processes and settings. The exam names AWS CloudFormation, which "enables you to create and provision AWS infrastructure deployments predictably and repeatedly."

7.2 Managed services absorb undifferentiated work

The PDF exam guide adds "Identifying managed AWS services (for example, Amazon RDS, Amazon ECS, Amazon EKS, Amazon DynamoDB)" to task 1.4. Each one moves work off your staff:

Fewer undifferentiated tasks means less staff time and less software to license: a direct line into the labour row of the TCO table in Section 6.5.

7.3 The shared-responsibility shift: EC2 vs RDS vs Lambda

The more managed the service, the smaller the customer's share of the stack. The line moves; it never disappears.

Constants that never move: physical security and the security of the cloud are always AWS; data, IAM and encryption decisions are always the customer. Shared controls, patch management, configuration management, awareness and training, are owned by each party for its own layer.

• 📚 Did you know? AWS's own phrasing matters: for abstracted services such as Amazon S3 and Amazon DynamoDB, "customers are responsible for managing their data (including encryption options) and using IAM tools to apply the appropriate permissions." And as infrastructure modernizes, "more responsibility is shifted onto the service provider"; which is a shift of the line, not a transfer of your data responsibility. Serverless is not "someone else's problem".

---

Real-World Case Studies

Every figure below is customer- or AWS-claimed and unaudited, attributed with its source so you can check it. The examinable point is always the pattern, which strategy was chosen, which lever moved, never the marketing number.

Capital One: "all in on AWS" (banking, regulated)

"We are truly all in on the cloud, and AWS has been instrumental in enabling us to take full advantage of the benefits of being in the cloud.": Chris Nims, SVP cloud and productivity engineering, Capital One (AWS case study)

Shutterfly / SBS: VMware to AWS (e-commerce printing)

"We've taken big steps to burn down technical debt so we can focus on transformation and move with the kind of agility we had in the early days.": Ian Wright, VP Infrastructure, Shutterfly/SBS (AWS case study)

• 📚 Did you know? A third migration pattern worth recognizing appears in the same research set: Bangkok Flight Services used AWS Application Migration Service (MGN) for a pure rehost, finishing in 7 months with "no rollback, no disruption", cutting IT infrastructure management time by 50% and replacing a missing DR site with a multi-AZ design; while Philip Morris International migrated 400 applications in two years and AWS's own write-up records that its initial aggressive lift-and-shift "yielded suboptimal results". Same R, opposite lesson: rehost is fast, but rehosting everything is a strategy error.

Netflix: global streaming at AWS scale (entertainment)

All Netflix figures above are AWS-published and accessed Oct 2026; anything about Netflix's pre-AWS outage history lives on Netflix's own site and is not taught here as AWS fact.

Box: Well-Architected cost optimization (enterprise SaaS)

"Our use of AWS best practices led to savings of over 2 million dollars, setting a new baseline…": Clay Alvord, director of FinOps and SRE, Box (AWS case study)

FarEye: Savings Plans + Spot + Graviton (SaaS logistics)

2026 Updates (as of October 2026)

Six sourced changes that touch migration and cloud economics since the last time you refreshed your notes: every figure date-stamped, and prices/limits drift, so verify current before use:

• Migration tooling changed scope. Migration Evaluator joined the in-scope service list (Migration and Transfer category) while AWS Snow Family moved to the out-of-scope list; in-scope entries went 128 → 111 and out-of-scope entries 11 → 55 (official CLF-C02 in-scope / out-of-scope pages, diff computed 2026-10). Both lists are explicitly "non-exhaustive and subject to change".

• A new commitment option arrived: Database Savings Plans, launched 2025-12-02 and advertised at up to 35% for a 1-year, no-upfront commitment (serverless up to 35%, provisioned up to 20%). Compute and EC2 Instance Savings Plans are unchanged, and the exam guide still names only "AWS Savings Plans" (AWS News Blog and What's New, 2025-12-02).

• The support lineup was rewritten on 2025-12-02: Business Support+ from 29 USD/month, AWS Enterprise at 5,000 USD/month and AWS Unified Operations at 50,000 USD/month, with Basic free; Developer, classic Business and Enterprise On-Ramp took no new subscriptions after 2025-12-02 (existing ones run through 2027-01-01), and Domain 4 Task 4.3 of the current exam guide now reads "Basic Support, AWS Business Support+, AWS Enterprise Support, AWS Unified Operations" (AWS News Blog, 2025-12-02; exam-guide PDF, © 2026; prices as of Oct 2026).

• Free Tier became account-age dependent (change effective 2025-07-15): accounts created on or after that date get 6 months (or until credits run out) plus USD 100 sign-up credit and up to USD 100 more, up to USD 200, on t3.micro, t3.small, t4g.micro, t4g.small, c7i-flex.large, m7i-flex.large; accounts created earlier keep the legacy 12-month tier with t2.micro / t3.micro (AWS News Blog, 2025-07-15; Amazon EC2 User Guide table "before and after July 15, 2025").

• Cost tooling keeps moving even though the examinable names do not: Cost Optimization Hub (2023-11-26), RI/SP Group Sharing generally available (2025-11-19), Target Coverage in Savings Plans Purchase Analyzer (2026-06-09) and the Well-Architected Agent preview (2026-10-01, which AWS calls "the next-gen evolution of AWS Trusted Advisor and the AWS Well-Architected Tool"). The exam still tests Cost Explorer, AWS Budgets, AWS Pricing Calculator, Trusted Advisor and the AWS Well-Architected Tool.

• Pricing and egress keep shifting under the TCO model: Amazon S3 Express One Zone cut storage about 31% and GET request prices up to 85% (effective 2025-04-10) and EC2 NVIDIA GPU instances up to 45% (2025-06-05); on the way out you still get 100 GB per month of free data transfer out from AWS Regions (a 2021 change, not a 2025 one) plus an approval-based 90-day window to finish moving data off AWS when you leave (AWS FAQs and blogs; as of Oct 2026).

• 📚 Did you know? Amazon S3's own price history is the cleanest published illustration of economies of scale: it launched on 14 March 2006 at $0.15 per GB per month serving "approximately one petabyte of total storage capacity across about 400 storage nodes", and AWS reported in March 2026 that it now charges "slightly over 2 cents per gigabyte": a reduction of approximately 85% since launch; while code written for S3 in 2006 "still works today, unchanged" (AWS News Blog, 2026-03-13; as of Oct 2026: verify current pricing before you quote it).

• 📚 Did you know? Some AWS cost stories are older than most candidates' study material: NASA reported saving "almost a million dollars in cost savings each year" on 2012-06-11, and NASA JPL now processes about 4.4 TB of downlinked data daily (generating up to 70 TB of final data products) using EC2 Auto Scaling with Spot Instances at up to a 90 percent discount compared to Amazon EC2 On-Demand pricing (AWS case study, accessed Oct 2026). Dates matter: never present a 2012 figure as a current one.

---

Practice Questions