Re: [Lsr] LSR: Using DSCP for path/topology selection Q

[Uma Chunduri <[e-mail]>]

A quick comment below –

Cheers & have a great weekend!

--
Uma C.


From: Lsr [mailto:[e-mail]] On Behalf Of Jeff Tantsura
Sent: Friday, November 16, 2018 12:15 AM
To: Robert Raszuk <[e-mail]>
Cc: [e-mail]; Les Ginsberg (ginsberg) <[e-mail]>; [e-mail]; [e-mail]; [e-mail]
Subject: Re: [Lsr] LSR: Using DSCP for path/topology selection Q

Robert,

match DSCP X
set context Y or plane Z doesn’t make it any different.
It has been used and abused in any possible way. If you want to write a BCP saying - use it for X/Y/Z but not for A/B/C because of.... - your business.

As to using it someplace else - I’d expect respective documents to cover the use, flex-algo drafts as to your example.

More fundamentally, (flex-algo is the best example) we have got context aware metadata in a form of: MPLS labels (SR SID), v6 EHs, plethora of overlay encaps, etc, with accompanying CP extensions that can be used to achieve exactly that.

Now tell me - why again DSCP?

P.S. in my previous life, working on 5G transport slicing (yes, i know :)) - i needed per slice identity over the common transport, we ended up looking at UDP port ranges, rather than DSCP - too few bits


[Uma]: +1 and was part of that previous life. May be it should be called as “4G slicing” and it’s deployed in few places.  Any ways it’s all about how to map the incoming traffic at PE (in the above case at CSR/P-GW) to a particular LSP. I didn’t see any requirement which needs an MTR/flex-algo aka light-weight-sub-topology which needs to be built based on DSCP.