Designing a web-facing network for redundancy and QoS (in a small business)

Question

Summary

How does one configure the web-facing part of a small business network for redundancy and quality of service? I mean ... we can have the control over our internal infrastructure, but what happens when it comes to the WAN links?

Context

• Small business with < 20 user devices connected and a small budget
• They have a LAN for users
• They also have a private network with some virtual servers, including a VoIP based PBX.
• Couple of servers will probably be migrated to the cloud in the future, but most of them will stay on-premise.
• They already have a firewall appliance that is used as a gateway for both private subnets.

Requirements

• They want two Internet uplinks for redundancy (on different technologies and/or different ISPs).
• They cannot afford enterprise grade connections, so the bandwidth of the two links can be susceptible to changes over time (as it happens for DSL connections). Also, the links will have different bandwidth one from each other. That said, they will be quite fast, quite stable and with a decent latency.
• They need public subnets for web-facing servers in their network. These subnets need to also work on failover links, so that their servers will be reachable by IPs even during the downtime of their main connection. This is especially critical for the VoIP system. I know that some ISPs provide public subnet preservation for failover links. I also know that sticking to a single provider would reduce redundancy in case of failure on the ISP side.
• They need QoS both for VoIP and for other classes of traffic.

Questions

1. How do you configure QoS in a situation where available bandwidth is unknown and/or can change over time?
2. How do you preserve subnets over multiple connections? Are there any low-budget alternatives to the single ISP solution?
3. If you go the single ISP way (that would preserve subnets), I would expect the ISP to either (A) provide a router that combines the two links or (B) provide two routers (one WAN link each) that will talk to each other using first-hop redundancy protocols. How can you configure QoS so that it works both for main and for secondary connection (that have different speeds)? How do you prevent double NAT, especially in the case B where you have first-hop redundancy?
4. How do people generally solve these types of challanges? Do they leave firewall and QoS configuration to the ISP to manage?

Answer 1

The bad news:

Without enterprise-grade circuits, you can't have QoS on the WAN. You can't control the QoS on the provider's network and they will very likely ignore any QoS signalling attempts you make.

The good news:

It often doesn't matter. If you've ever used a product like Google Voice, MS Teams, FaceTime, Facebook Messenger, etc. You know that voice quality on the Internet is acceptable most of the time without QoS. Any periods of poor voice quality are likely to be transitory.

There are many ways of handling failover with multiple ISPs, but they all involve additional equipment/expense that you may or may not be willing to incur. If you're planning a migration to the cloud, you might consider a load balancer-as-a-service to route connections to your active circuit.