Please. Looking for examples of Hierarchical QoS (HQoS) keyed towards provisioning of Customer Edge (CE) ROUTERS facing carrier Provider Edge (PE) devices of an MPLS network.
Specifically Cisco has moved through some design changes and marketing name changes for the 'Modular QoS CLI' (MQC) and added features, which have their respective examples, but the examples are autonomous or a la carte and not shown as they'd be when integrated into a more comprehensive solution scheme.
For example a 5 class traffic model for a CE router connected to an MPLS network via Ethernet and ATM (DSL) to the PE utilizing Committed Access Rates in EF, AF4x, AF3x, AF2x, AF1x and best effort or default class traffic AND leveraging the Cisco 'Ethernet Overhead Accounting' that includes a real world example e.g. Carriers charge for all the overhead:
• Pre-amble 7
• Start-of-Frame delimiter 1
• MAC destination 6
• MAC source 6
• VLAN TAG 4
• Ethertype/Length 2
• Payload 46-1500
• CRC32 4
• Interframe Gap 12
not just the MACs, ethertype, payload and CRC's; the model(s) should include a variety of classification methods to highlight the Cisco software including policing of the EF traffic where exceeding traffic is dropped and another where it's marked down; Utilization of Cisco's Low Latency Queuing (LLQ) for that higher traffic; other classes should include newer Cisco shaping techniques like Class Based Shaping instead of the Generic Traffic Shaping usually found; drop management techniques should be included in the bottom classes to further highlight the software's modular nature like Distributed Weighted Random Early Detection (dWRED).
To add real value and really show off Cisco's software instead of the answer to this question being a few different examples of the QoS classification, marking and queuing in the middle of a generic configuration, it should also add other 'feature flash' (I'll call it) like adjustment of Cisco's queue-limit; setting of the hardware transmit ring; use and effect of the interface bandwidth command (both inside and outside of the policy-map's classes); max-reserved-bandwidth modifications and effect; advantages in using Cisco's random-detect exponential-weighting-constant value and so on.
It's great to have all these features, Cisco has great software. However I've found that mixing the features in a design at your desk is difficult. In order to create a good design it's necessary to be sitting at the CLI to test the limits of integration since the parameters and their variable bounds change based on context. What may appear to work on paper based on all the documentation, is either rejected by the IOS during implementation, or accepted by the IOS but has no effect, or empirically demonstrates a completely unexpected behavior. Having examples that Cisco marketing and engineering has put some serious thought into in terms of showing off feature integration, that are complete tuned configurations top to bottom, would be incredibly valuable to those whom are able extrapolate to derive answers to their configuration challenges through inferred information in those feature uses and boundaries.
I hope you're able to use this, or maybe poll to validate my assertions first and then, hopefully, execute. Or maybe you'll just send me a link that I've been missing all along – THAT would be great!
cheers,
->N
