Bgp?

[SteveA]

I need a dynamic routing capability that can failover when a link goes down. I can tweak rip to do it, but the downtime is 20+ seconds. OSPF has had such mixed reviews on the WAR platform that I've been unwilling to invest the time on something smarter people than me have been unable to fix

OLSR description sounds like just the right thing, but after reading the WIKI and following the links provided to read what the curent authors say about it, OLSR is not suitable for a commercial grade connection. It sounds like they have given up on it


From https://www.open-mesh.net/optimized-link-state-routing-deamon/olsr-story.txt/view
* Deployment of olsr-0.4.10

Results:
Now it is really working and usable :)

It's still not absolutely loop-free

Multihop-Links with 10 Hops work and are stable as long
as the wireless links work.

LinkQualityDjikstraLimit allows to run olsr even on a
realtively slow CPU in a big Mesh-Cloud -
but the routing-table becomes very very static

Gateway-Switching is still a constant annoyance if a mesh
has more than one Internet-Gateway

Conclusions:
Apart from the problems with Gateway-Switching it is now a
well behaving routing protocol


But still... Thomas and I agreed that we could cope with the increasing size
of the Freifunk-Networks only by making the protocol more and more static.
Link State Routing has significant design flaws. Why does every node
calculate a whole routing table from every node to every node - if all it
can do is decide which direct neighbor it sends the packet to? Synchronized
Link State Information is impossible to achieve in a wireless network. Why
let every CPU calculate unneccessary information? Link State Routing thinks
too much and is far too complex for is own good. Why do all this?


We decided to come up with something better. Thomas had the idea for a name:
B.A.T.M.A.N - Better Approach To Mobile Ad-Hoc Networking.
Any chance of BGP coming back? This was discussed many months ago as a maybe, but I have heard nothing about BGP for awhile.

[greg]

BGP is in v2 and it worked well for me on a PC platform. Are you talking on a core router or at your AP's?

[tony]

We will bring BGP to the desktop edition in one of the upcoming releases.

[tog]

But no iBGP for WARs?

Not that I would need it, I use OLSR and it works great for me anyway.

[tony]

Due to the space constraints of the Embedded platforms, there will be no BGP available for them (at least, not at the moment).

[oscarBravo]

I'm not convinced BGP is a good fit for SteveA's requirements anyway - it converges slowly, by design.

[lonnie]

SteveA, I read that article and far from giving up they pretty much wrote code and made it work. The proplems they had were in previous releases and the summary at the end states that 0.4.10 indeed works and i usable, as long as the wireless links work. DUH. This simply means that you, as the RF designer, are responsible for building a solid RF link. I'm sorry but I don't see that as a shortcoming of OLSR. It will work around a total failure of a link but a constantly flipping out link can give trouble, but then that causes every sort of routing and even bridging mechanism to have trouble.

There is no magic. Build a solid link and it'll work as well as you hoped. Take some shortcuts hoping that a special routing protocol will fix it and you will unhappy.

[Physcon]

We are still using OSPF on our network and it actually works pretty well most of the time. I say most of the time because it will still occasionally do strange things where the neighbors will see each other but not exchange routes. However, most of the time things run just fine and if it does happen "freak out", disabling, and re-enabling OSPF on one or both sides usually fixes the problem. Whats really made it work best for us is by turning off broadcast updates and manually specifying the neighbors. That seems to make things more reliable.

Currently, our network has 9 backhauls in a ring-like configuration for redundant routes and if we ever take down a link for whatever reason OSPF has always re-routed things just like its supposed to do. All the problems with OSPF we've seen just seem to happen randomly or sometimes when a staros router gets rebooted.

Overall i still like OSPF on staros, its just a little psycho sometimes. Its definitely a better solution for us than oslr because we have to interface our network with other types of routers and Cisco's just dont speak oslr. Besides, i like to stay a little more compatable than using a proprietary routing protocol on my network.

Oh and also, all of our backhaul WARs are running 1.1.4 or earlier. We even have a few v2 routers out there still and they talk to the v3 routers just fine.

[SteveA]

Thanks everyone for the responces.

The convergence time of BGP is slow, but it works eventually. If not an option though we'll have to look at the other options more closely.

The latest authors of OLSR mentioned that "Gateway-Switching is still a constant annoyance if a mesh has more than one Internet-Gateway". I have more than one internet gateway, and having the system reconverge on a different gateway would be a problem. I could force everything through a single gateway then let it decide on the routing, but my gateways are located at different physical locations.

There were lots of complaints previously posted about OLSR and OSPF, but I have used neither on WARs yet. Is the community consensus that either will work now?

thanks...

Steve

[tog]

OLSR has been doing great for me for around a year now.