What is TE really for?

[jlixfeld]

So I have TE configured on an EoMPLS VPN, but it doesn't seem to be throttling anything.

Traffic is only throttled if I put a service-policy on the PE subinterface.

When you set the ip rsvp bandwidth on the PE-PE interfaces and specify tunnel mpls traffic-eng bandwidth on the tunnel, isn't traffic supposed to get throttled when the traffic on the tunnel exceeds what's configured by the tunnel bandwidth command?

Am I missing something?

[mheusing · Accepted solution]

Hi,

MPLS TE is a control plane feature, i.e. there is no impact on packet handling besides path selection. From a QoS perspective pure MPLS TE does not make any difference.

MPLS TE, in simple words, is a technology to replace destination based path selection with a more flexible mechanism taking into account more parameters than just IGP metric.

With MPLS TE RSVP is used to signal the selected path, but different to IntServ, it does not modify the QoS behaviour.

So if you need QoS you need service policies whether or not MPLS TE is in place.

Now in a further step DiffServ-aware MPLS TE was introduced. In combination with proper QoS policies this allows you to give hard SLAs f.e. for VoIP traffic.

So MPLS TE is used to efficiently map traffic onto available ressources and with Fast Reroute to overcome f.e. link failures in less than 50 ms.

For further details you might want to read the MPLS TE section in the MPLS configuration guide for your IOS/platform.

In addition there are several Books on MPLS TE or you read the respective RFCs dealing with MPLS TE.

Hope this helps!

Regards, Martin

[jlixfeld]

So if I have a single link between PEs, TE is useless. QoS is helpful, but TE is useless. Yes?

[mheusing]

Hi,

yes, TE gets interesting, when there is a choice for the path.

QoS is helpful or even required to meet SLAs for certain traffic f.e. VoIP. It will have to be configured on a per link/interface basis.

Regards, Martin

[jlixfeld]

Hmm, then in that case can I use TE on the single path and then QoS on the xconnect subinterface, or do I still need to do QoS on the physical links along the path between PEs?

[mheusing]

Hi,

QoS implementedwith the widely used DiffServ model, treats resources independently. There is no QoS signalling involved in DiffServ. Thus every interface along the whole traffic path, which potentially can be congested needs to be configured separately. Do not forget potential backup pathes in your QoS design.

In other words: for a customer it is pretty irrelevant, WHERE a packet is dropped in the SP network - gone is gone. Thus the SP has to make sure on every interface- from input through core to output - that the SLAs are met.

In simple words:

MPLS TE is a complex way of doing IP routing, nothing more and nothing less. So MPLS TE helps you to direct the traffic to a desired path, QoS will make sure that important traffic is treated properly along this path.

You might need to implement both technologies to meet your service offerings.

Hope this helps!

Regards, Martin

[rajju]

Hi Martin:

Thank you for your excellent reply.

Learning a lot from you, Harold, Swaroop and the other guys here in this forum.

Looks like now you are with Cisco ;->

Sincerely.

[mheusing]

Hi Niraj,

You are welcome! Yes, I am with Cisco now, delivering training through AS Education.

Regards, Martin