Hybrid Connectivity Deep Dive¶
Connecting on-premises networks to cloud environments.
VPN Types¶
flowchart LR
subgraph ONP[On-prem]
HOST[Workloads]
GW[Customer gateway<br/>router / firewall]
end
subgraph OPTIONS[Connectivity choices]
S2S[Site-to-site VPN<br/>IPsec over internet<br/>~1-1.25 Gbps per tunnel]
CV[Client VPN<br/>per-user remote access]
DX[Direct Connect /<br/>ExpressRoute /<br/>Interconnect<br/>1-100 Gbps dedicated]
end
subgraph CLD[Cloud]
VPN_GW[VPN gateway /<br/>VPN concentrator]
DX_GW[DX gateway /<br/>ER circuit termination]
VPC[(VPC / VNet)]
VPN_GW --> VPC
DX_GW --> VPC
end
GW --> S2S --> VPN_GW
CV --> VPN_GW
GW --> DX --> DX_GW
HOST --> GW Site-to-Site VPN¶
A persistent, encrypted tunnel between your on-premises network and a cloud VPC/VNet.
How It Works - Uses IPsec (Internet Protocol Security) to encrypt traffic over the public internet - Requires a customer gateway device (router or firewall) on-premises - Creates two tunnels per connection for redundancy (AWS), or a configurable number of tunnels (Azure, GCP)
AWS Site-to-Site VPN
On-premises <--IPsec--> Virtual Private Gateway (or Transit Gateway) <--> VPC
Azure VPN Gateway
On-premises <--IPsec--> Azure VPN Gateway <--> VNet
GCP Cloud VPN
On-premises <--IPsec--> Cloud VPN Gateway <--> VPC
Client VPN (Remote Access)¶
Allows individual users or devices to connect to cloud resources.
AWS Client VPN - OpenVPN-based managed service - Supports Active Directory and SAML-based authentication - Split tunnel or full tunnel routing - Docs: https://docs.aws.amazon.com/vpn/latest/clientvpn-admin/what-is.html
Azure Point-to-Site VPN - Supports OpenVPN, SSTP, and IKEv2 protocols - Certificate-based or Entra ID authentication - Integrated with Azure VPN Gateway - Docs: https://learn.microsoft.com/en-us/azure/vpn-gateway/point-to-site-about
GCP Identity-Aware Proxy (IAP) - No VPN required - tunnels SSH and TCP traffic through Google's network - Uses identity-based access instead of network-level access - Alternative to traditional client VPN for many use cases - Docs: https://cloud.google.com/iap/docs/concepts-overview
Dedicated Interconnects¶
AWS Direct Connect¶
Overview - Dedicated 1 Gbps, 10 Gbps, or 100 Gbps connections at Direct Connect locations - Hosted connections available from 50 Mbps to 10 Gbps through partners - Lower latency and more consistent performance than VPN
Architecture
On-premises <--> Colocation facility (Direct Connect location) <--> AWS Region
Key Concepts - Virtual Interfaces (VIFs): - Private VIF - connects to a VPC via Virtual Private Gateway - Public VIF - connects to AWS public services (S3, DynamoDB, etc.) - Transit VIF - connects to a Transit Gateway for multi-VPC access - LAG (Link Aggregation Group) - bundle multiple connections for increased bandwidth - Direct Connect Gateway - connect to VPCs in multiple regions from a single connection
Docs: https://docs.aws.amazon.com/directconnect/latest/UserGuide/Welcome.html
Azure ExpressRoute¶
Overview - Dedicated connections via connectivity providers (50 Mbps to 10 Gbps standard, 100 Gbps Direct) - ExpressRoute Direct provides dedicated port pairs (10 Gbps or 100 Gbps) - Private peering for VNet access, Microsoft peering for Microsoft 365 and Azure public services
Key Concepts - Peering types: - Private peering - connects to VNets - Microsoft peering - connects to Microsoft 365 and Azure PaaS services - ExpressRoute Global Reach - connect on-premises sites through Microsoft backbone - ExpressRoute FastPath - bypasses the gateway for improved data path performance - Redundancy - two circuits per peering location by default
Docs: https://learn.microsoft.com/en-us/azure/expressroute/expressroute-introduction
GCP Cloud Interconnect¶
Overview - Dedicated Interconnect: 10 Gbps or 100 Gbps connections at Google peering edges - Partner Interconnect: 50 Mbps to 50 Gbps through service providers - No encryption by default (traffic stays on private network)
Key Concepts - VLAN attachments connect the interconnect to VPC networks - Up to 16 VLAN attachments per interconnect - Can use Cloud Router for dynamic routing with BGP - Cross-Cloud Interconnect connects directly to other cloud providers
Docs: https://cloud.google.com/network-connectivity/docs/interconnect/concepts/overview
Transit Architectures¶
AWS Transit Gateway¶
Architecture
VPC-A --|
VPC-B --|-- Transit Gateway --|--> VPN to on-premises
VPC-C --| |--> Direct Connect
Key Features - Hub-and-spoke connectivity for multiple VPCs and on-premises networks - Supports up to 5,000 attachments - Route tables for network segmentation - Inter-region peering between Transit Gateways - Multicast support - 50 Gbps bandwidth per VPC attachment
Docs: https://docs.aws.amazon.com/vpc/latest/tgw/what-is-transit-gateway.html
Azure Virtual WAN¶
Architecture
VNet-A --|
VNet-B --|-- Virtual WAN Hub --|--> VPN to on-premises
VNet-C --| |--> ExpressRoute
Key Features - Managed hub-and-spoke with automatic routing - Supports VPN, ExpressRoute, and point-to-site in the same hub - Hub-to-hub connectivity across regions - Integrated with Azure Firewall and third-party NVAs - Routing intent for simplified security policies
Docs: https://learn.microsoft.com/en-us/azure/virtual-wan/virtual-wan-about
GCP Cloud Router¶
Key Features - Provides dynamic BGP routing for VPN and Interconnect - Advertises VPC subnet routes to on-premises automatically - Custom route advertisements for specific prefixes - Works with HA VPN and Cloud Interconnect
Network Connectivity Center - Hub-and-spoke topology for hybrid connectivity - Supports multiple spokes (VPN, Interconnect, Router Appliance) - Cross-cloud and multi-cloud networking
Docs: https://cloud.google.com/network-connectivity/docs/router/concepts/overview
Redundancy Patterns¶
Dual Tunnel VPN¶
- Every VPN connection should use at least two tunnels
- Tunnels should terminate on different devices or availability zones
- Use BGP for automatic failover when a tunnel goes down
- Monitor tunnel status and alert on failures
Diverse Provider Redundancy¶
For Direct Connect / ExpressRoute / Interconnect 1. Single connection - no redundancy (not recommended for production) 2. Two connections at the same location - protects against device failure 3. Two connections at different locations - protects against facility failure 4. Two connections from different providers - protects against provider failure
Recommended Production Setup
On-premises --> Provider A (Location 1) --> Cloud
On-premises --> Provider B (Location 2) --> Cloud
On-premises --> VPN (backup) -----------> Cloud
VPN as Backup for Dedicated Connections¶
- Configure VPN with lower BGP priority than the dedicated connection
- Traffic automatically fails over to VPN if the dedicated link goes down
- VPN provides lower bandwidth but is quick to provision
- Test failover regularly
Bandwidth and Latency Considerations¶
Bandwidth Comparison¶
| Method | Bandwidth | Typical Latency | Cost Model |
|---|---|---|---|
| Site-to-Site VPN | 1.25-3 Gbps per tunnel | Variable (internet) | Per hour + data transfer |
| Direct Connect / ExpressRoute / Interconnect | 1-100 Gbps | Low, consistent | Port hour + data transfer |
| Hosted / Partner connections | 50 Mbps - 50 Gbps | Low, consistent | Provider pricing |
Latency Factors¶
- Geographic distance between on-premises and cloud region
- Number of network hops
- VPN encryption/decryption overhead (5-15% CPU overhead)
- Dedicated connections eliminate internet routing variability
- Use cloud backbone for inter-region traffic when possible
Data Transfer Costs¶
- VPN: standard data transfer out charges apply
- Direct Connect / ExpressRoute: reduced data transfer rates
- Ingress is generally free, egress is charged
- Consider data transfer costs when choosing regions and connectivity methods
Right-Sizing Guidance¶
- Start with VPN for proof of concept and low-bandwidth workloads
- Move to dedicated connections when you need consistent latency or bandwidth above 1 Gbps
- Use multiple connections for high-availability production workloads
- Consider SD-WAN overlay for application-level traffic steering