TGW vs VPC Peering
Scale considerations.
Overview
Transit Gateway vs VPC Peering chooses AWS VPC connectivity. Peering is simple and cheap for pairs; TGW scales to many VPCs through a hub but costs per-attachment plus per-byte. The discipline matches the right tool to VPC count.
- Scale considerations. Per-VPC-count the right choice; the math flips from peering to TGW around five-to-ten VPCs.
- VPC peering: simple. One-to-one connection; matches small topologies with few VPCs.
- Transit Gateway: hub-and-spoke. Many VPCs through central hub; scales to dozens of VPCs without quadratic peering complexity.
- Peering no transitive routing plus TGW cost. Per-pair peering connections cannot transit; TGW costs per-attachment plus per-byte.
The approach
The practical approach: peering for two VPCs, TGW for many across regions, inter-region TGW peering for global hub-and-spoke, monitor cross-VPC bandwidth, documented per-VPC topology. The team’s discipline produces matched connectivity.
- Peering for pairs. Two VPCs that need direct connectivity; cheap and simple at small scale.
- Transit Gateway for many. Multiple VPCs across multiple regions; scales without quadratic peering complexity.
- Inter-region peering for TGW. Connect TGWs across regions; produces global hub-and-spoke.
- Monitor cross-VPC bandwidth plus documented topology. Cost surveillance catches surprises; per-VPC connectivity committed for investigation.
Why this compounds
The discipline compounds across architecture decisions. Each correct choice produces ongoing value; the team’s networking expertise grows; new VPCs inherit the connectivity pattern.
- Better resilience. Cross-region connectivity supports DR; the architecture survives single-region failure.
- Better global architecture. Workloads near users with shared services; produces fast user experience.
- Cost predictability. Monitored bandwidth produces predictable costs; supports finance conversations.
- Institutional knowledge. Each connection teaches networking; the team’s networking muscle grows.
TGW vs peering is an infrastructure decision that pays off across years. Nova AI Ops integrates with networking telemetry, surfaces patterns, and supports the team’s connectivity discipline.