Common Network Designs

Exam guide§2.3

Two recurring patterns for making an application more resilient: spread it across zones for availability, then across regions for globalization.

Increased availability with multiple zones

Place VMs in different zones of the same region, on a single subnet. A zone is an isolated failure domain, so a zonal outage takes down one VM, not both.

Two VMs, two zones, one subnet (10.2.0.0/16) - resilient to a zonal failure with no extra security config.

#NumbersWhy a single subnet

• A subnet is regional - it spans every zone in its region, so both zonal VMs share the range 10.2.0.0/16.
• One subnet means one firewall rule covers both VMs - availability without added security complexity.
• A regional managed instance group does this for you: it spreads instances across zones in a region automatically.

Globalization with multiple regions

Put resources in different regions. Regions are independent of each other's infrastructure, hardware, and software failures, so this is a stronger isolation than zones alone.

Resources in separate regions with their own subnets - the highest degree of failure independence.

#NumbersWhat multiple regions buy you

• Failure independence: a region-wide outage still leaves the other region serving.
• Each region has its own subnet range (10.2.0.0/16 vs 192.168.0.0/16).
• A global external Application Load Balancer routes each user to the closest region - better latency and lower network egress cost.

⚠GotchaZones protect against hardware, regions protect against a region

Multi-zone gives you isolation from most hardware/software failures within a region. Only multi-region survives a whole-region failure. Managed instance groups + global load balancing are the standard building blocks for both.