問題文のキーワードとサービス संकेतを見つける

試験でキーワードが重要な理由

コストを示すキーワード

# Cost keyword -> answer mapping:
# 'lowest cost for batch jobs' -> Spot Instances
# 'archive rarely accessed data' -> S3 Glacier Deep Archive
# 'steady workload for 3 years' -> Reserved Instances or Savings Plans
# 'event-driven, unpredictable traffic' -> Lambda (pay per invocation)
# 'reduce inter-Region data transfer' -> CloudFront or regional caching
# 'remove idle EC2 instances' -> AWS Compute Optimizer right-sizing

パフォーマンスを示すキーワード

# Performance keyword -> answer mapping:
# 'sub-millisecond cache' -> ElastiCache (Redis)
# 'global static content delivery' -> CloudFront
# 'HPC / tightly coupled compute' -> Cluster placement group, EFA
# 'shared POSIX filesystem across Linux instances' -> Amazon EFS
# 'Windows SMB file shares' -> FSx for Windows File Server
# 'in-memory key-value, session state' -> ElastiCache (Memcached or Redis)
# 'relational + scales to millions of writes' -> Aurora

セキュリティを示すキーワード

# Security keyword -> answer mapping:
# 'IAM role for EC2 to access S3' -> EC2 instance profile with IAM role
# 'prevent public access to S3 bucket' -> Block Public Access + bucket policy
# 'audit all API calls in account' -> AWS CloudTrail
# 'ensure resources comply with policy' -> AWS Config + Config Rules
# 'TLS certificate for ALB' -> AWS Certificate Manager (ACM)
# 'cross-account role assumption' -> IAM Trust Policy + STS AssumeRole

運用上のオーバーヘッドを示すキーワード

# 'least operational overhead' -> managed service wins
# Self-managed on EC2          -> Managed AWS equivalent
# MySQL on EC2                 -> Amazon RDS for MySQL
# Redis on EC2                 -> Amazon ElastiCache for Redis
# Kafka on EC2                 -> Amazon MSK (Managed Streaming for Kafka)
# Elasticsearch on EC2         -> Amazon OpenSearch Service
# Custom batch scheduler       -> AWS Batch
# Custom ETL scripts on EC2    -> AWS Glue (serverless ETL)

スケーラビリティと伸縮性を示すキーワード

# Scalability keyword -> answer mapping:
# 'scale to zero when no traffic' -> Lambda or Fargate (task count 0)
# 'handle flash sales / burst traffic' -> ALB + ASG with step scaling
# 'unpredictable database read/write' -> DynamoDB On-Demand mode
# 'stream processing with variable throughput' -> Kinesis On-Demand
# 'global traffic with automatic failover' -> Route 53 health check failover
# 'scale read capacity for RDS' -> Read Replicas (up to 15 for Aurora)

移行とモダナイゼーションを示すキーワード

# Migration keyword -> answer mapping:
# 'petabytes of data, slow internet' -> AWS Snowball Edge (offline transfer)
# 'replicate servers to AWS with < 1 hour RTO' -> AWS MGN
# 'Oracle to PostgreSQL schema conversion' -> AWS SCT + DMS
# 'ongoing DB replication, minimal downtime' -> DMS Full Load + CDC
# 'inventory 5000 on-premises VMs' -> Application Discovery Service
# 'assess portfolio and recommend strategy' -> Migration Hub Strategy Recommendations

データベース選定キーワード

# Database keyword -> correct service:
# 'ACID transactions, joins' -> RDS (MySQL/PostgreSQL) or Aurora
# 'single-digit millisecond at any scale' -> DynamoDB
# 'session caching, Pub/Sub, leaderboard' -> ElastiCache Redis
# 'complex SQL analytics, BI' -> Amazon Redshift
# 'social network, fraud detection graphs' -> Amazon Neptune
# 'IoT sensor time series' -> Amazon Timestream
# 'document / JSON flexibility' -> DynamoDB or DocumentDB

疎結合化とメッセージングのキーワード

# Decoupling keyword -> service:
# 'buffer requests, async processing' -> SQS Standard
# 'order must be preserved, no duplicates' -> SQS FIFO
# 'notify many services of one event' -> SNS Topic
# 'react to AWS service state changes' -> EventBridge (default bus)
# 'stream analytics data to S3 automatically' -> Kinesis Firehose
# 'multiple apps read same data stream independently' -> Kinesis Data Streams
# 'custom event routing by message content' -> EventBridge with event patterns

サーバーレスアーキテクチャのキーワード

# Serverless keyword -> service:
# 'code triggered by events, no EC2' -> Lambda
# 'REST API with no infrastructure' -> API Gateway + Lambda
# 'containers without cluster management' -> ECS Fargate
# 'Kubernetes without node management' -> EKS Fargate Profiles
# 'SQL queries on S3, no warehouse' -> Amazon Athena
# 'NoSQL database, no capacity management' -> DynamoDB On-Demand
# 'workflow orchestration, no servers' -> AWS Step Functions
# 'ETL without Spark cluster' -> AWS Glue (serverless Spark)

ネットワークと接続のキーワード

# Networking keyword -> service mapping:
# 'private dedicated bandwidth to AWS' -> AWS Direct Connect
# 'encrypted tunnel over internet to AWS' -> Site-to-Site VPN
# 'accelerate global TCP/UDP traffic' -> AWS Global Accelerator
# 'connect 100 VPCs without full mesh' -> AWS Transit Gateway
# 'private access to S3 from VPC' -> S3 Gateway Endpoint (free)
# 'private access to SQS/SNS from VPC' -> Interface Endpoint (PrivateLink)
# 'accelerate static content globally' -> CloudFront CDN

クイックチェック

レッスンのまとめ