Welcome to the Microsoft Q&A Platform. Thank you for reaching out & I hope you are doing well.
I understand that you would like to know the best practices for deploying a Site-to-Site connection to connect your OnPrem and Azure resources.
The above architecture is not something we would recommend.
Our recommendation is to go with either
- Hub-Spoke Architecture and use only one VPN Gateway to connect to OnPrem
- or
- Connect your sites in different geographic regions to your OnPrem individually and do not use VNET Peering at all
Technically speaking,
- In your architecture, if one S2S Fails (let's say East US), it will not be able to connect to OnPrem using other S2S connections
- This is an expected behavior and we cannot over ride this.
- The reason is because
- A VNET can use only one VPN Gateway - either it's own or the one to which it's peered.
- It cannot use both at any instant
- A VNET can use only one VPN Gateway - either it's own or the one to which it's peered.
- So, you can have a Hub Spoke as below (no VNET Peering between Spokes and Spokes are only peered to Hub)
- Or, below - No peering between the Hubs
Now, I am not sure why you would need a Load Balancing here.
If you are concerned about the throughput and availability of the Tunnel, you should consider,
1.Zone-redundant virtual network gateway
- This is for availability
- Deploying gateways in Azure availability zones physically and logically separates gateways within a region, while protecting your on-premises network connectivity to Azure from zone-level failures
- The two gateway instances will be deployed in any 2 out of the three Availability zones to provide zone-redundancy.
- Refer : https://learn.microsoft.com/en-us/azure/vpn-gateway/about-zone-redundant-vnet-gateways
2.Throughput
- Gateway SKUs by tunnel, connection, and throughput : https://learn.microsoft.com/en-us/azure/vpn-gateway/vpn-gateway-about-vpngateways#benchmark
- Consider using Active-Active Gateways : https://learn.microsoft.com/en-us/azure/vpn-gateway/vpn-gateway-highlyavailable#dual-redundancy-active-active-vpn-gateways-for-both-azure-and-on-premises-networks
- This can create upto 4 Tunnel in a single VPN Connection
- This makes sure even if one tunnel goes down, the rest of the 3 remains active
- Also, if an instance goes down in one availability zone, another instance in another availability zone will continue to keep 2 remaining connections active
- Also, if an instance goes down in one availability zone, another instance in another availability zone will continue to keep 2 remaining connections active
- This makes sure even if one tunnel goes down, the rest of the 3 remains active
- This can create upto 4 Tunnel in a single VPN Connection
Load Balancing and High availability options are as mentioned above for a VPN Site to Site.
While we cannot directly call this Load Balancing, you can see how it helps to have traffic in 2 Connections and there by, 4 tunnels (2 Tunnel in each connection) helps to improve throughput.
P.S:
- I would say the best scenario for your requirement is to go with Azure vWAN with Branch to Branch enabled
- Refer : https://learn.microsoft.com/en-us/azure/virtual-wan/virtual-wan-global-transit-network-architecture#hubtohub
- Here, everything is connected via vWAN even across regions.
Kindly let us know if this helps or you need further assistance on this issue.
Thanks,
Kapil
Please don’t forget to close the thread by clicking "Accept the answer" wherever the information provided helps you, as this can be beneficial to other community members.