MAM and RDM: control-plane and data-plane differences?

[GIANRICO FICHERA]

MAM and RDM are control plane features and in some cases I find very difficult to find the operational differences when I have a real configuration with data plane traffic. From a theoretical point of view all is very easy. From a purely pratical point of view is difficult to build a QoS configuration with MAM model without wondering if you could do the same thing with RDM but with a different QoS configuration.

Cisco IOS supports two BC: BC0 (global pool) and BC1 (sub-pool). Let's imagine to have two MPLS-TE tunnel with RSVP-TE reservation. With RDM if there is space left from BC1, at that time BC0 is able to use that bandwidth. With MAM even in the case there is no traffic in BC1, BC0 is unable to use that bandwidth.

My question is: do not depend on data-plane QoS configuration? If I have a MPLS-TE network with DS-TE and RDM in many cases I use rate-limit or bandwidth or priority commands to avoid that a tunnel occupies more bandwith than reserved. From an ISP point of view this is normal because a tunnel can be a leased line for a customer. First question:  What does that mean? That I have MAM in my network because tunnels respect the RSVP reservation?

As the second example, I configure RSVP with MAM in my network. But if I don't use commands like rate-limit or bandwidth or priority what prevents BC0 to take BC1 bandwidth if available? But this is a RDM behaviour. So I have both RDM and MAM in my network?

Thank you in advance for your help

Gianrico Fichera
ITESYS SRL

[Vasilii Mikhailovskii]

Hello.

I understand that it's late to answer, but I will put some information here if anybody else would search for similar topic.

So, MAM and RDM are described by RFCs 4125 and 4127 and they are about bandwidth reservation/allocation.

The primary difference is described in the following article - http://www.cisco.com/c/en/us/td/docs/ios/12_2sr/12_2srb/feature/guide/dsteietf.html (Table 1).

RDM without pre-emption would suffer if class 7 allocates all the bandwidth before class 0 is signalled.

MAM allows class isolation without pre-emption, as you may restrict classes to some bandwidth allocation independently of each other.

 Usage Guidelines

1. The Maximum Allocation Model should be selected when the network administrator needs to ensure isolation across all Class Types without having to use pre-emption, and can afford to risk some QoS degradation of Class Types other than the Preimum Class.

2. The Russian Dolls Model should be selected when the network administrator needs to prevent QoS degradation of all Class Types and can impose pre-emption. 

---

So, both models are about control-plane, not data-plane, - we can't say, that policing is a part of any of them, or if it may replace one. But, policing is /may be a part of an approach for deploying "guaranteed" service (as described in DS-TE RFC 3564).