backup routing network

[Christoph Zach]

hello

We have 2 locations, A and B
At location A we have an 5308XL
At Location B we have an 2626

The 2 locations are connected through an Laserlink 100MBit and as backup an wireless bridge 24 Mbit.


To be able to route between these 2 locations
we will make 2 routingnetworks
10.100.0.0/24 (Laser) and
10.100.1.0/24 (wireless bridge)

We only have static routes because the 2626 dont have Routing Protocols like OSPF or RIP

WE want the 10.100.0.0 the primary routingnet and the 10.100.1.0 the backup routing net, so when the laser fails, (possible because of Fog)
we switch over to the routingnet 10.100.1.0 with the wireless Bridge

But the 5308 XL and the 2626 dont support metrics for static routes.
How can i do that ???

[Matt Hobbs]

As you're already aware the 2626 doesn't support RIP or OSPF to which you can assign a path cost / metric. Because of this and the limitations with static routes, I don't think you'll have any luck with layer 3 protocols.

What I'm hoping you may be able to do instead is just use spanning-tree, to which you can definitely assign a path cost or priority if required.

I'd be interested to find out how you go.

[Christoph Zach]

Hi Matt
Thanks for answer

Alternatives would be
a.) by a second 5300xl for Location B
b.) use only one routingnet an use rstp for switching bestween laser and wireless

Problem ist, we use VOIP over this Link, and RSTP needs 2 seconds to switch over, its quite fast, but to slow for VOIP.

I think with ospf and redundant routingnetworks it will be faster, or ??

How long will ospf need to switch between 2 routingnetworks ?

[Matt Hobbs]

That's a good question and something that I don't have that much experience with (OSPF). Looking through the manual, I'm guessing that the OSPF hello interval would need to be set the lowest value of 1 second. The dead interval would also need to be decreased (4?).

In this type of setup where the link is most likely to be lost between the wireless devices, the switches themselves are going to keep the physical link up to it's end of the wireless device, that's why I think the hello interval would be important in this situation. You're relying on a layer 3 failover.

If the actual port on the 5300xl went down, then it would be much more immediate and could probably failover quicker (layer 2).

As I mentioned I am guessing here, but it is something I'd like to test out myself in the near future.

What I would do to test is put a 5300xl on each end, and in between I'll put 4 switches, 2 for each VLAN. I will then test failover by pulling the link between the middle two switches, and once again by pulling the link that goes directly into a 5300xl.

Cisco support Fast Hello's which look like they'd be helpful for your setup:

http://www.cisco.com/univercd/cc/td/doc/product/software/ios120/120newft/120limit/120s/120s23/fasthelo.htm

[Mohieddin Kharnoub]

Hi

Matt, Reducing the Hello interval to 1 second is a good idea in terms of convergence because that what we are looking for, but we have 2 Issues here :

1- The Network will be flooded with OSPF multicast every 1 Sec, and since there is no DR, so every OSPF aware router will do this flood.

2- Reducing Hello Interval to minimum used in LAN to get Maximum Convergence.

In our case we have a WAN based Network and i don't think you Christoph will get less than 2 Seconds in case of failover.

I would recommend you go for RSTP since you already have the equiepments and try it, and later on you can decide which solution you have to go for.

Its good from you to share us whats going on.

Good Luck !!!