WAN Redundancy

[pschwalger]

Between two datacenter sites we have a WAN connection provided by a local Telco and routing between sites is using BGP.

We also have a layer 2 fibre end to end connection between the datacenters for diversity, although this fibre link is not part of the BGP routing process.

I would like to enable automated failover of the routing between sites utilizing the secondary end to end fibre connection in the event that we lose the the primary connection.

Currently, I am only utilizing the end to end fibre connection for some traffic using policy based routing. In the event of a BGP outage I have to manually add static routing via the secondary fibre link to reconnect the sites.

I would be interested in how I could better automate this process and utilize both circuits for redundancy, bgp, ip sla's?.

I would appreciate any recommendations or direction on a possible solution.

thanks, Peter.

[rothomas2]

First option is to roll out Dynamic routing.  Dynamic routing always beats IP SLA. You should failover to use IP SLA for things like full reachability over a path that you couldnt otherwise with Hellos from a Dynamic routing protocol like OSPF. 

Second, although you could use BGP and the TCP hellos this will probably end up with slow failover. This might be ok if you have almost no outages, but generally you want something faster like IS-IS or OSPF. 

I really like using OSPF Fast Hellos. If you got two good fibers, OSPF Fast Hellos you can rival Subsecond failover time from other Technologies like SDH. 

[Akash Agrawal]

Hi Peter,

You can bundle links in to port-channel so that on layer 3 router will not realize even when one link going down and traffic getting switchover to other link.

Second thing you can run any IGP protocol over links and increase cost on one which will act as back link.

If you want to improve convergence time, you can use BFD for fast failure detection

Please don't forget to rate this post if it has been helpful

Regards,

Akash

[pschwalger]

Thanks guys for your recommendations...I just want to get a little more detail to move forward

We have a pretty minimal setup with a router at each datacenter. The current primary connection currently runs E-BGP via a local Telco and most of our traffic is routed this way.

A second Telco then provided a layer 2 fiber connection which we have terminated at both ends on the datacenter routers as a /30 end to end directly connected connection.

Both connections are approx 20Mb metro between the datacenters with 100Mb access connections terminations.

I would like to utilize both links as much as I can and use dynamic routing to have routes fail over if one of the links goes down.

As mentioned, I only currently use the direct fiber connection by implementing policy based routing and if the primary E-BGP connection goes down I have to add static routing to use the direct router connection.

I don't really care which connection is the primary / secondary but I do want to use both links and have dynamic failover.

Would you recommend declaring an internal routing process for the direct fibre connection using OSPF, RIP v2 or EIGRP for this?

If the traffic then prefers the direct fiber link for routing the datacenter to datacenter traffic is it possible to use Policy based routing across the E-BGP link?

Any other things to watch out or be careful with this set up?

I know that we should have dual routers etc at each end, but we currently just have a warm spare at each end and I want to move forward from the static routing arrangement and then tackle the single point of failure issue a little bit later.

thanks, Peter.