Answer the question
In order to leave comments, you need to log in
How to build a seamless transition between Internet access points?
There is a physical server, which requires constant uninterrupted access to the Internet. It is required to exclude situations when, for any reason, access to the Internet is lost. To do this, I want to use two access points - the main wired provider and the Internet from a flash drive. How to ensure the transition between access points without interrupting Internet access (even for a second) if access point 1 stops working?
Answer the question
In order to leave comments, you need to log in
no way.
This can be solved at the level of your own autonomous system and full-fledged bgp routing. But this is completely different money, knowledge, equipment.
Place the server on a colocation in a large DC with several channels. and then it won't be your problem.
It's actually very simple (in theory):
Set up two network interfaces on the server and set up your service to listen on both interfaces.
Then rewrite the client so that it uses both server interfaces and, if there is no response from the main one, resends the request to the backup
one. You can also not transfer the fault tolerance layer to the application level, but choose a ready-made transport layer protocol that will provide the necessary functionality, for example SCTP. True, in this case, you will also have to rewrite the server :)
As for the proposed options with BGP, it must be borne in mind that the standard holdtime for BGP on the Internet is 180 seconds. And for highways tier-1,2 many times more. There can be no talk of any subsecond switching via BGP
Well, with an adult approach, not only the network is reserved, but also the servers themselves. Collect them in pools and balance the load with load balancers (LB). At the same time, the issue of fault tolerance creeps to the level of these LBs themselves. LB fault tolerance is provided by a powerful infrastructure of cloud and/or service providers: AWS/GCP/Azure/CloudFlare
You should understand that when switching between providers occurs, then connections - first of all tcp they will still be broken and re-established, since your server will change IP address and route, how programs / services running on your server will react to this - you more visible.
Usually, uninterrupted access to the Internet is solved by installing a server in a data center with an appropriate level of redundancy, starting from two network cards assembled in a bond on the server, then two switches in the stack on the rack, and so on ....
Since you did not specify the OS on the server, then there is no possibility to advise you on the implementation of switching between providers.
> constant uninterrupted access to the Internet.
Only SD-WAN solutions are suitable for such requirements. All other solutions will make some kind of delay in switching and TCP break.
If you agree to put up with a switching delay, then
- rent a VPS
- buy a router, better mikrotik
- build two VPN channels between the VPS and the router
- run a dynamic routing protocol inside the VPN
- configure NAT on the VPS
With this scheme, TCP will not break, and the switching will be equal to rebuilding speed of the dynamic routing protocol.
Didn't find what you were looking for?
Ask your questionAsk a Question
731 491 924 answers to any question