Increasing the availability of services from the Internet?

Question

High availability

There is a small website and several online services.
Unfortunately, taking into account their specifics, it is not possible to take them to hosting providers.
And also, taking into account the region, there are problems at the level of providers in the region (they do not provide 100% availability).
But at the same time, there have not yet been cases that several providers are unavailable.
Question:
What are the options to ensure the availability of the site from the Internet through several providers, so that if one provider has problems, the service will automatically open through another provider.
PS. At the service level, this cannot be implemented, because. we cannot change it.
P.S.S. All providers have a white external IP address.
PSSS. A solution with minimal or no financial costs is desirable)

Answer 1

You buy an inexpensive vds-ku from a reliable provider, you put nginx there.
At home you play two, three, N servers, you set up replication.
Nginx on vds-ke proxies requests to the main server, in case of its fall - to another.

Answer 2

Unfortunately, taking into account their specifics, it is not possible to take them to hosting providers.

The previous discussion came down to VPS hosting providers…
Do you have a server at home (office)? Do all provider channels lead to it?
1. For example, we put Mikrotik. We connect all providers and set up a channel transfer when it falls to the worker. I don’t know about other manufacturers, but Mikrotik has implemented this very well. At least do all 5-10 ports with different providers.
If DNS round-robin is not suitable, you can file the same DDNS. And when changing the provider, run the script for changing the active ip.
2. Another option is a third-party service (the same ping admin, for example, or make your own analogue) that will monitor the availability of your service by ip. If the active one is dead, switch to another SP. But in this case, it will be difficult to monitor a 1-2 minute downtime…
3. We take any reliable hosting / VPS and proxy through nginx
upstream backend {
server ip1;
server ip2;
server ip3;
…
}
But in this case there will be delays depending on where to get this VPS…

Answer 3

You really suffer from garbage, in fact, this is easily provided by the same bgp, but you have everything at home, take a cheap hosting and don't worry.
Really the problem is not worth a damn.

Answer 4

Managing incoming traffic is a rather interesting and not simple thing.
IMHO: The simplest and most reliable solution is its own AS + BGP. In fact, the channels to the operators are Directs through which the BGP neighborhood rises. And the servers have an IP from your range. Thus, with the existing scheme, you get redundancy not to the server, but to the network. In normal mode, your IPs are available through all providers (well, or how you set it up), and in the event of a fall of some, through the survivors. At the same time, the server IP does not change, DNS does not need to be corrected, the switching occurs quite quickly (from personal experience - from 30 seconds to 1-2 minutes)

Answer 5

it all depends on the size of the network (number of IPs) and the pliability of providers.
we have PA addresses (/26), so we pay nothing for them, but pay for the BGP support service and the Internet. The final price tag for all services provided is about 30-40 thousand per month