Load Balancing Blitz: Google Cloud Load Balancing Introduction
💡 Originally published on Google Cloud Community on Medium.
TL;DR
Google Cloud Load Balancing (GCLB) is a software-defined, fully managed distributed routing platform built on Google's Maglev and Andromeda SDN network infrastructure. Unlike traditional hardware appliances or virtual machine appliances, Cloud Load Balancing requires no pre-warming, scales instantaneously to millions of queries per second, and provides a single global Anycast IP address that routes users to the closest available compute backend across the globe.
The Load Balancer Decision Matrix
graph TD
A["Incoming Traffic Profile"] --> B{"Is traffic External (Internet) or Internal (VPC)?"}
B -- "External (Public Internet)" --> C{"What protocol is used?"}
C -- "HTTP / HTTPS / gRPC" --> D["Global External Application Load Balancer (ALB)"]
C -- "TCP / UDP (Layer 4)" --> E["External Network Load Balancer (NLB)"]
B -- "Internal (Private VPC)" --> F{"What protocol is used?"}
F -- "HTTP / HTTPS (Microservices)" --> G["Internal Application Load Balancer"]
F -- "TCP / UDP (Databases / RPC)" --> H["Internal Passthrough Network Load Balancer"]1. Global vs Regional Load Balancers
- Global Anycast Load Balancing: Uses a single virtual IP (VIP) advertised from all Google edge points-of-presence (PoPs). User traffic enters Google's private backbone network at the closest geographical edge and travels on dedicated fiber directly to the application backend.
- Regional Load Balancing: Confines traffic routing within a single Google Cloud region for compliance, data sovereignty, or low-latency local VPC architectures.
2. Layer 7 (Application) vs Layer 4 (Network) Load Balancing
| Feature | Application Load Balancer (L7) | Network Load Balancer (L4) |
|---|---|---|
| OSI Layer | Layer 7 (HTTP, HTTPS, HTTP/2, gRPC) | Layer 4 (TCP, UDP, ESP) |
| Routing Intelligence | URL path matching, host headers, query parameters | IP address, port, and protocol tuple |
| SSL / TLS Termination | Terminated at edge with Google-managed certificates | Passthrough or proxy modes |
| Cloud Armor Security | Native DDoS & WAF rule integration | DDoS defense via Cloud Armor network edge |
| Backend Types | GKE Ingress, Cloud Run, Compute Engine MIGs, Cloud Storage | Compute Engine VMs, GKE nodes |
3. Health Checks and Failover
Cloud Load Balancers continuously monitor backend availability using configurable health checks:
- Proactive Removal: If a backend instance fails health probes, traffic is rerouted within milliseconds with zero dropped connections.
- Cross-Region Overflow: If the primary region reaches $100\%$ capacity utilization, the Global Application Load Balancer automatically spills excess traffic over to healthy backends in the next-closest region.