SDWAN QoS and Per-Tunnel QoS issue

[John.lc]

HI experts

I have three questions:

1. Does Queue0 have a recommended minimum bandwidth ratio in QoS settings?(If there is a 100Mbps link, what should we set the bandwidth ratio of Queue0)?
2. Whether Per-Tunnel QoS support for full-mesh topology, there is a one site as a important node
3. Can the bandwidth of the hub node of Per-Tunnel QoS be less than the sum of the downstream bandwidth configurations of Spokes?

For example, the Hub Site has a bandwidth of 1Gbps, and there are 10 Spoke sites below, and each site sets its downstream bandwidth as:
Site1: 400Mbps
Site2,3,4: 200Mbps
Site5: 500Mbps
Site6,7: 100Mbps
SIte8,9,10: 300Mbps

[John.lc]

Oh, no one knows about this problem?

[Kanan Huseynli]

Hi,

1) I don't think that there is standard way for bandwidth ratio for Q0 (LLQ). Maybe TAC can recommend some value

2) Per-tunnel QoS is supported only in hub&spoke topology

3) Downstream bandwidth by spoke is needed to not load spoke device. So, this configuration will work, but hub may sent only based on its configured shaping value, but not based on total value of spoke downstream bandwidths.

This document describes per-qos configuration and limitation. However, there is no more details in this document.

https://www.cisco.com/c/en/us/td/docs/routers/sdwan/configuration/qos/ios-xe-17/qos-book-xe/per-tunnel-qos.html

[John.lc]

HI Kanan,

Thanks for your response.

I read the documentation you provided before asking the question, but I still don't quite understand it.
According to what you mean, if I have 20 sites, even if the downstream bandwidth of each of my Spoke sites is set to 500Mbps, and the Hub node only has a bandwidth of 1Gbps, this will not cause traffic preemption on the Hub, right? In addition, if these 20 sites download a large file at the same time, will the 1Gbps bandwidth of the Hub be roughly divided into 20 parts, and at most each Spoke site will have a bandwidth of 50Mbps from the Hub?

[Kanan Huseynli]

Hi,

as you also understand, 500Mbps means that hub will not sent traffic more than 500m to spoke. On the other hand, 1G on hub means that, hub may send at most 1G traffic per sec.

Regarding your question, to be honest I don't quite sure. Logically, yes it should divide into 20 parts since there is no other value to differentiate "spoke" sites. However, maybe hub will send more traffic to spoke 1, for example. Document says that "The amount of outbound hub bandwidth that a “greedy” spoke can consume can be limited; therefore, the traffic can’t monopolize a hub’s resources and starve other spokes. " - but we don't know how it is implemented. If you have TAC access ask as a question.

There is need to do lab /test to understand fully how it works.

[John.lc]

Thank you for your reply again, this question is really confusing, thank you anyway.