Cisco Support Community
cancel
Showing results for 
Search instead for 
Did you mean: 
New Member

convergence time less 1s

Hello,

Is there a "routing" protocol existing which can garantee a convergence time (swapping from working to backup path) in less then 1s. Or can we fine tune hello timers to achieve this?

4 REPLIES
New Member

Re: convergence time less 1s

Both OSPF and EIGRP can provide pretty low convergence times, though I don't think it's possible to trim it down to 1sec.

OSPF send hello packets, after not receiving 4 of those it marks a router as dead. The hello timer can be set to 1 second, so you could achieve convergence of about 4-5 seconds.

You have to consider what type of connection you have too.. is this over the Internet, if so, packet loss could result in a router getting marked as dead.

You would better have a fast network for this.

Gold

Re: convergence time less 1s

In recent IOS code, you can configure OSPF to mark neighbors down in about one second, with support for sending hello packets at subsecond intervals.

http://www.cisco.com/en/US/products/sw/iosswrel/ps1839/products_feature_guide09186a00801039b1.html

I don't know of any similar enhancements for EIGRP.

Bronze

Re: convergence time less 1s

ISIS has had this functionality for a good while in Cisco code. OSPF, as mentioned, just got it recently.

New Member

Re: convergence time less 1s

You can fine tune EIGRP using offset-lists and adjusting the delay on various interfaces. If you make your adjustments such that your backup path becomes a feasible successor, you can achieve sub-second convergence when you primary path fails. If EIGRP has to query for a route, you are dead in the water.

Doing this gives you the benefits of a dynamic routing protocol, while keeping things very deterministic in the case of failure (you know exactly which secondary path is going to pick up the traffic) and making failover as fast as possible.

Figuring out what to use for your offset and delay values takes a little bit of calculating (depending how complex your topology is), but nothing too hard (you just have to make sure your primary path wins by at least 1).

s.

84
Views
0
Helpful
4
Replies
CreatePlease to create content