Compute Services

Service	Model	You manage	When to use	Scale/Cost feel
Compute Engine (GCE)	IaaS	OS, patching, networking, autoscaling groups	Lift-and-shift VMs, custom images, special kernels	Full control ↔ more ops
Google Kubernetes Engine (GKE)	CaaS	Pods/Deployments; Google manages control plane (mostly)	Microservices, portability, orchestration	Scales pods/nodes; pay for nodes & control plane tier
App Engine	PaaS	앱 코드만 (runtime 제한)	웹앱, API with minimal ops	Autoscale; per-use billing
Cloud Run	Serverless containers	Container image; infra = managed	Stateless HTTP/event apps, internal services	Scale to 0; pay per request / CPU-RAM-time (100ms granularity)
Cloud Functions (2nd gen on Cloud Run)	FaaS	Function code + triggers	Event-driven, lightweight APIs, schedulers	Scale per-invocation; pay only while running

Compute Engine: “VMs on demand”, full flexibility, instance templates + MIGs.
GKE: Container orchestration; portability; Kubernetes primitives.
App Engine: Fully managed PaaS; focus on code; standard vs flexible runtimes.
Cloud Run: Knative 기반, stateless containers, scale to zero, request or Pub/Sub triggers, run anywhere Knative runs.
Cloud Functions: Events → function 실행; 초단위 과금; 2nd gen은 Cloud Run infra 덕에 콜드스타트/네트워킹 개선.

Virtual Private Cloud (VPC)

Core objects

Projects, Networks (VPC), Subnetworks (regional), Regions/Zones, IP addresses, VMs, Routes, Firewall rules

Projects · Networks · Subnetworks

Project = 리소스/서비스/빌링의 최상위 컨테이너.
VPC Network = global scope (전 세계 region에 걸침). IP range 자체는 subnet에 있음; network는 논리 컨테이너.
Subnetwork = regional. 같은 region 내 여러 zones를 가로지름.

VPC types

Default: 프로젝트 생성 시 자동. 각 region에 미리 만든 /20 subnets, 기본 firewall(ingress ICMP/RDP/SSH from anywhere, 그리고 intra-VPC all protocols).
Auto mode: 모든 region에 1개 subnet 자동 생성. CIDR은 10.128.0.0/9 범위에서 할당. /20 → /16까지 확장 가능.
Custom mode: 수동으로 subnet/IP 설계. 서로 겹치면 안 됨.
- Auto → Custom 전환은 one-way (되돌릴 수 없음).
- IPv6 dual-stack 지원(예: VM에 IPv4+IPv6).

Network isolation & global reach

같은 VPC면 region 달라도 내부 IP로 통신(구글 글로벌 백본).