소거법 및 오답 패턴

소거가 필수적인 이유

오답 패턴 1: 잘못된 서비스 범주

# Example distractor elimination by service category:
# Question: Reduce read latency for an RDS MySQL database
# A. Enable RDS Multi-AZ -> NOT for read performance (it is for HA)
# B. Add an ElastiCache Redis read-through cache -> CORRECT category
# C. Enable AWS CloudTrail on the RDS instance -> WRONG category (audit)
# D. Increase EC2 instance type of the app servers -> WRONG layer (app, not DB)
#
# Eliminate A, C, D by category/purpose mismatch. Answer: B

오답 패턴 2: 올바른 서비스, 잘못된 계층

# Example tier distractor:
# Question: Most cost-effective storage for compliance backups accessed once every 5 years
# A. S3 Standard -> too expensive for cold data
# B. S3 Standard-IA -> lower cost but still too expensive for rarely accessed
# C. S3 Glacier Instant Retrieval -> millisecond retrieval (not needed here)
# D. S3 Glacier Deep Archive -> CORRECT: lowest cost, 12-hr retrieval OK for 5yr access
#
# Key: match access frequency to the right storage class tier

오답 패턴 3: 잘못된 scope

# Scope distractor examples:
# Q: Protect against AZ failure for a web tier
# Wrong scope: Route 53 latency routing across Regions (that is Region-level)
# Correct scope: Multi-AZ ASG with ALB spanning multiple AZs in one Region
#
# Q: Enforce a policy that prevents any IAM user from disabling CloudTrail
# Wrong scope: S3 bucket policy on the CloudTrail bucket
# Correct scope: SCP (Service Control Policy) in AWS Organizations

오답 패턴 4: 기술적으로 유효하지만 지나치게 복잡함

# Over-engineering distractors:
# Q: Shared POSIX file system for 50 EC2 Linux instances
# Distractor: Set up an EC2 NFS server with EBS volumes (works but complex)
# Correct: Amazon EFS (managed, auto-scaling, no servers to manage)
#
# Q: Managed Kubernetes without node management
# Distractor: EKS with self-managed node groups (you manage EC2 patching)
# Correct: EKS with Fargate profiles (no nodes to manage)
#
# Q: Trigger Lambda on new S3 object
# Distractor: Use EventBridge with a custom polling Lambda that scans S3
# Correct: S3 Event Notification directly to Lambda (simpler, lower latency)

오답 패턴 5: 일관성 모델 불일치

# Consistency/durability distractor:
# Q: Process financial transactions in strict order, no duplicates
# Distractor: SQS Standard -> at-least-once, no guaranteed order
# Correct: SQS FIFO -> exactly-once, message order per group
#
# Q: Multiple consumers must independently replay the last 7 days of events
# Distractor: SNS -> no storage, no replay
# Distractor: SQS -> messages deleted after consumption
# Correct: Kinesis Data Streams with 7-day retention -> supports replay

두 단계 풀이 방법

# Two-pass elimination template:
# Pass 1 - eliminate obvious violations:
#   Does A use the wrong AWS service? -> Eliminate
#   Does B violate 'no server management'? -> Eliminate
#   Does C apply to the wrong scope? -> Eliminate
#   D: remaining candidate
#
# Pass 2 - apply primary constraint:
#   Primary constraint: 'lowest cost'
#   D vs. remaining: which is cheaper given the access pattern?
#   -> Final answer selected

답을 바꾸는 한정 표현

# Qualifying word examples:
# 'protect data WITHOUT encrypting existing objects' -> S3 default encryption
#   (only encrypts new objects, not retroactively) + separate migration step
#
# 'AUTOMATICALLY remediate non-compliant resources' -> AWS Config + SSM Remediation
#   (not just Config rules which only detect, but not auto-fix)
#
# 'EXISTING monolith, move with NO code changes' -> Rehost / Lift-and-shift / MGN
#   (eliminates Refactor/Lambda/microservices options)

'Best Practice' 문제 다루기

# Best practice distractor examples:
# Q: Best practice for granting an EC2 instance access to S3
# A. Embed IAM access keys in the application code -> NEVER (credential risk)
# B. Store IAM credentials in EC2 environment variables -> still exposed
# C. Attach an IAM instance profile (role) to the EC2 instance -> CORRECT
# D. Create a shared IAM user for all EC2 instances -> violates least-privilege
#
# Eliminate A, B, D by security best practice. Answer: C

연습: 여러 제약 조건에 따른 소거

# Multi-constraint elimination walkthrough:
# Q: 'Serverless + highly available + REST API + most cost-effective + variable traffic'
#
# Option A: ALB + EC2 ASG with On-Demand -> fails 'serverless'
# Option B: API Gateway + Lambda (no reserved concurrency) -> satisfies ALL constraints
# Option C: API Gateway + EC2 backend -> fails 'serverless'
# Option D: API Gateway + Lambda with MAX reserved concurrency -> fails 'cost-effective'
#            (reserved concurrency charges even when idle in some pricing models)
#
# -> Answer: B

독해 함정: NOT 및 EXCEPT

# Examples of negatively phrased questions:
# 'Which of the following is NOT supported by AWS DMS as a migration source?'
# -> You need the one service DMS cannot read FROM
#
# 'Which statement about SQS FIFO queues is INCORRECT?'
# -> Three answers are true; one is false
#
# 'Which option provides the LEAST operational overhead?'
# -> NOT asking for most overhead; pick the most automated/managed
#
# Tip: if you see NOT/EXCEPT/INCORRECT, immediately re-read all options
# as potential TRUE statements and identify the outlier

빠른 확인

lesson 복습