When dscp or cos pass through Nexus7K

[wfqk]

Hi Please see below diagram

SW1 ---- Nexus7K ----SW2

We setup marking on traffic at SW1 while there are no any Qos configuration on Nexus7K. The question is if SW2 can see the same marking when SW2 receive the traffic? Thank you

[Peter Paluch]

Hello,

Without additional configuration to explicitly remark packets or frames, Nexus switches are completely transparent to IP DSCP and 802.1Q CoS markings and do not touch them in any way. What comes in also goes out. Of course, CoS markings can only be retained if the egress port is a trunk, otherwise the VLAN tag including the CoS bits would be lost.

Please feel welcome to ask further!

Best regards,
Peter

[wfqk]

Thank you so much for your reply!

It looks like Nexus7K has different action at this point with general old switch such as 3560 etc, which can rewrite these marking. That also means we do not need to do any configurations, including trust configuration etc on that Nexus7K port if we want to send traffic with the marking over to SW2? 

" --- CoS markings can only be retained if the egress port is a trunk, otherwise the VLAN tag including the CoS bits would be lost ---" that means it is only on trunk, not access port. Why is that? Is this because the marking is attached with trunk frame? 

[Peter Paluch]

Hello,

You are very much welcome!

It looks like Nexus7K has different action at this point with general old switch such as 3560 etc, which can rewrite these marking.

Yes, that is entirely correct. As opposed to Catalyst switches that had the concept of "trust cos", "trust dscp", and then you had the cos-dscp and dscp-cos rewrite maps, there no such thing on Nexus switches, neither the concept of trust, nor the concept of rewriting one of CoS or DSCP values depending on the trusted one.

That also means we do not need to do any configurations, including trust configuration etc on that Nexus7K port if we want to send traffic with the marking over to SW2?

Absolutely correct.

" --- CoS markings can only be retained if the egress port is a trunk, otherwise the VLAN tag including the CoS bits would be lost ---" that means it is only on trunk, not access port. Why is that? Is this because the marking is attached with trunk frame? 

Again, absolutely correct! CoS bits are a part of the 802.1Q VLAN tag. Plain Ethernet frame format does not contain any field for CoS marking; it has to be VLAN-tagged to have the CoS field. Since access ports do not use VLAN tagging, there can be no CoS bits present on access ports.

Feel welcome to ask further!

Best regards,
Peter

[Joseph W. Doherty]

BTW, and to add to Peter's information, traditionally Cisco switches like the 3560 wouldn't require you to "trust" until you enabled QoS. I.e. they too left ToS or CoS alone until they were configured to "do something" about QoS. (Also, if they were not configured to "trust" or have an ingress policy, they would remark the QoS tag to zero.)

Current gen Cisco switches (or "older" gen with current gen sups and IOS), now tend to behave more like Cisco routers regarding QoS. I.e. The ToS/CoS tags aren't changed unless a QoS policy is configured to do so.

The bottom line, though, is you often need to check what any individual Cisco switch will do.

Also BTW, some of the Nexus series have, shall we say, "interesting" QoS features.

[wfqk]

Thank you!!