GCP Introduction

Infra layers

• Base: Networking & Security, foundation for everything
• Middle: Compute & Storage, decoupled so they scale independently
• Top: Big Data & ML, ingest → store → process → deliver insights/pipelines/models
• Managed infra → infra scaling

Course focus

• Focus = Compute/Storage + Big Data/ML
• Networking/Security = out of scope

Compute

Big Data는 분산 병렬 처리가 기본 방식 Cheap storage > multiple computers (compute) > combine (reduce)

Locations & structure

• 5 geos: North America, South America, Europe, Asia, Australia

  → 위치 선택이 availability/durability/latency에 직결. (지연 = 출발→도착 왕복 시간)

• Regions & Zones

  • Region = independent geo area; multiple zones inside

    예: europe-west2 (London) = 3 zones

  • Zone = where resources actually run. Zonal resource는 해당 zone 장애 시 영향

• Placement scopes: Zonal / Regional / Multi-Regional 선택 가능

  • 장애 도메인 분리 + 사용자 근접성(latency) 최적화

• Multi-Region services 예시

  • Spanner multi-region: replicas across zones & regions (예: NL + BE) → low-latency reads 근접 지역에서 가능

Compute portfolio

• Compute Engine (IaaS)
  • VM 중심. you manage instances/storage/networking
• GKE (Google Kubernetes Engine)
  • Containers on managed Kubernetes (vs individual VMs)
• App Engine (PaaS)
  • Fully managed app platform. 코드 + 라이브러리 바인딩. Infra abstracted → 앱 로직에 집중