MPLS TE tunnel metric

[dknov]

Hi,

Let's say you have configured a Tunnel interface to calculate its path using TE metric (not IGP) by using "tunnel mpls traffic-eng path-selection metric te" command. Now tunnel headend router needs to know what is a cost (metric) to send traffic via this tunnel. Does it use TE metric or IGP metric?

If it uses TE metric, when router needs to consider path through TE tunnel and a path through normal IGP, does it consider IGP cost and TE cost to be 1:1 comparable, so cost of 10 in TE metric equals cost 10 in IGP metric?

And another question, what if you do not configure TE metric on all interfaces, but then you configure Tunnel interface to calculate the path using TE metric. Does this mean that router will only consider interfaces on which TE metric is configured, or will it use all interfaces and use IGP metric "as if" it was TE metric?

Thanks,

David

[Harold Ritter]

The "tunnel mpls traffic-eng path-selection metric te" will influence the path that the tunnel selects. Once the tunnel is setup, the metric for the best IGP path between tunnel headend and tailend is used to decide whether data should be forwarded through the tunnel or not.

If "mpls traffic-eng administrative-weight" is not configured on a given link, the CSPF takes these links into accounts but uses the IGP metric whether "tunnel mpls traffic-eng path-selection metric" specifies IGP or TE.

Hope this helps,

[dknov]

Excellent, thanks!

Now few follow-up questions.

1. As far as I understand TE tunnel is no different from GRE tunnel as far as routing selection is concerned. This means that if prefix A is known via both IGP path and TE autoroute, router will do a normal next hop selection based on best cost to reach A, am I correct? This would mean that you need to carefully plan the cost through TE tunnel to make it prefered and there will be no "special" preference for the TE tunnel, right?

2. Now I would like to understand load-balancing with equal and unequal costs.

a. Is it correct to say that with unequal cost router will make a selection based on least cost regardless if this is for TE tunnel tailend or any prefix "behind" it, meaning traffic might take TE or IGP path based on best metric?

b. Is it correct to say that with equal cost for the tunnel tailend, router will always prefer TE path rather than IGP path?

c. Is it correct to say that when there are two TE tunnels with different bandwidth requirements (and having the best path cost), router will automatically loadbalance between the two tunnels?

This means that cost needs to be the same, but bandwidth requirements might differ.

Thanks,

David

[Harold Ritter]

1. It is a bit more complex than that. If traffic is destined to the tunnel tailend then it will be forwarded down the TE tunnel even if you have an equal IGP metric via another path.

If the traffic is to a destination downstream from the tunnel tailend and you have an alternate path with an equal metric then traffic will be loadshared.

2.

a. The cost of the TE tunnel is always the shortest IGP metric to the TE tailend regardless which path is taken.

b. correct

c. if the two TE tunnel are to the same tailend but via different paths, traffic will indeed e loadshared between the two tunnels.

Hope thsi helps,

[dknov]

Hi,

How about if you have IGP path to tailend that has a cost of 20 and TE tunnel cost that is 25 (you have it manually configured with "tunnel mpls traffic-eng autoroute metric" command).

Will router choose to forward traffic towards the tailend IP through IGP rather than TE tunnel?

Thanks,

David

[Harold Ritter]

By default, the path weight for the TE tunnel would be 20 if the best IGP path is 20 and traffic would then be send down the tunnel.

If you manually change the Tunnel path weight with the "tunnel mpls traffic-eng autoroute metric 25" then the traffic would use the IGP path rather than the TE Tunnel.

Hope this helps,

[dknov]

Hi,

Can the following scenario happen?

Headend router is configured to use TE weight for best path calculation, so path is chosen, however since TE tunnel cost is in fact the IGP cost and not TE weight (per your previous answers) traffic might never actually use the tunnel because other IGP costs might be lower.

This scenario is possible due to a fact that different metric types are use to establish path and assign cost to it.

With all this said, the solution of using TE weight for path calculation might actually backfire at you, since you might be creating tunnels that you'll never pass traffic through.

David

[Harold Ritter]

As I stated in my last email, unless you change the metric, the best IGP metric is used and traffic to the tunnel tailend will definitely go down the tunnel.

If you manually change the TE tunnel metric to a value higher than the best IGP metric to the tunnel tailend, the traffic will be forwarded using the IGP path.

So it is indeed possible to have a situation where the tunnel is up and no traffic if forwarded through it.

Hope this helps,