Hi Ali and Menachem, thank you for the discussion of the applicability of PW CW. I would like to bring to your attention the work at the MPLS WG on the use of the Post-stack First Nibble (PFN). <https://www.ietf.org/archive/id/draft-ietf-mpls-1stnibble-02.txt> I must apologize that the draft has lapsed. The authors are finalizing updates, and the new version will be uploaded before IETF-119. It seems like one of the key updates in this draft, the intention to deprecate and ultimately obsolete cases where a non-IP payload encapsulated in MPLS without the presence of a Post-Stack Header (e.g., PW CW) is relevant to this discussion. I've asked for a presentation slot at the BESS WG session in Brisbane to share the update on this work.
Regards, Greg On Mon, Feb 5, 2024 at 10:06 AM Ali Sajassi (sajassi) <sajassi= 40cisco....@dmarc.ietf.org> wrote: > Hi Menachem, > > > > The use of control word is not mandatory and it is situation dependent. > Both RFC 7432 (and now bis) and RFC 8469 (which is basically elaboration of > section 18 of RFC7432/bis) mention that the control word is not needed when > there is no chance of packet re-ordering – e.g., when underlay tunnel is > RSVP-TE. Also, when the network (inclusive of all PE and P nodes) uses > Entropy Label, then there is no chance of re-ordering either. So, we are > just saying that in scenarios where there is no chance of packet > re-ordering, then control word is not needed (to avoid packet re-ordering) > – i.e. no need to tax the packet with additional 4 bytes. > > > > So, I was suggesting the text to be clarified as follow: > > > > - If a network (inclusive of both PE and P nodes) uses entropy labels > per [RFC6790] for ECMP load balancing, then the control word MAY NOT be > used. > > > > This means if the operators still want to use the control word with EL, > then they still can! > > > > Cheers, > > Ali > > > > > > *From: *Menachem Dodge <mdo...@drivenets.com> > *Date: *Monday, February 5, 2024 at 5:55 AM > *To: *Ali Sajassi (sajassi) <saja...@cisco.com>, Matthew Bocci (Nokia) > <matthew.bocci=40nokia....@dmarc.ietf.org>, > draft-ietf-bess-rfc7432...@ietf.org <draft-ietf-bess-rfc7432...@ietf.org> > *Cc: *bess-cha...@ietf.org <bess-cha...@ietf.org>, bess@ietf.org < > bess@ietf.org> > *Subject: *Re: Mail regarding draft-ietf-bess-rfc7432bis > > Hello Ali, > > > > Thank you kindly for your response. > > > > The question that Mathew and I raised, is why make the control-word > dependent on the presence of the Entropy Label (per RFC6790)? > > > > Transit Routers may or may not perform their load balancing based on the > Entropy Label. > > Some transit routers do perform deep packet inspection whether or not the > Entropy Label is present (whether or not it is needed), > in which case the presence of the control-word is important. > > > > Why not let the network administrator decide whether a control-word should > be present? > > > > Mathew wrote as follows, see also that the CW can be included for > additional reasons and the reference to RFC8649: > > “*The head end PE has no idea what hashing mechanism is actually used > downstream, regardless of whether the entropy label is inserted by it. The > entropy label is just there to provide additional flow information if the > downstream P router is load balancing based on the label stack, but it does > not in itself prevent the P router from scanning below the bottom of stack > and instead load balancing on the payload after checking the MPLS first > nibble. This also seems to be superseded by RFC8469 and all the discussion > over the years about making CW mandatory for MPLS-based services . It is > also worth noting that CW is not just to prevent aliasing between IP and > Ethernet traffic, but can be used to indicate OAM or other types of > maintenance packets**.”* > > > > So, we were suggesting that the text be removed, to remove the dependency > between the Entropy label and the control-word. > > > > And then, we would need an errata for RFC 8214 to remove the following text: > > “If a network uses entropy labels per [RFC6790 > <https://datatracker.ietf.org/doc/html/rfc6790>], then the C Flag > > MUST NOT be set, and the control word MUST NOT be used when sending > EVPN-encapsulated packets over a P2P LSP.” > > > > Appreciate your inputs in understanding if there is indeed a reason for > the dependency between the Entropy Label (per RFC6790) and the CW. > > > > Thank you kindly. > > > > Best Regards, > > Menachem > > > > > > *From: *Ali Sajassi (sajassi) <saja...@cisco.com> > *Date: *Monday, 5 February 2024 at 7:52 > *To: *Menachem Dodge <mdo...@drivenets.com>, Matthew Bocci (Nokia) > <matthew.bocci=40nokia....@dmarc.ietf.org>, > draft-ietf-bess-rfc7432...@ietf.org <draft-ietf-bess-rfc7432...@ietf.org> > *Cc: *bess-cha...@ietf.org <bess-cha...@ietf.org>, bess@ietf.org < > bess@ietf.org> > *Subject: *Re: Mail regarding draft-ietf-bess-rfc7432bis > > CAUTION: External E-Mail - Use caution with links and attachments > > > > Hi Matthew, Menachem: > > > > The text in the yellow says: “If a network uses entropy labels per > [RFC6790]” … > > It should be noted that the word “network” is used which is inclusive of > all the PE and P nodes in that network. So, if the network uses entropy > labels and does ECMP based on that, then there shouldn’t be a need for > control word. However, I don’t mind changing it from “SHOULD NOT” to “MAY > NOT”. > > > > Cheers, > > Ali > > > > *From: *Menachem Dodge <mdo...@drivenets.com> > *Date: *Sunday, February 4, 2024 at 12:39 AM > *To: *Matthew Bocci (Nokia) <matthew.bocci=40nokia....@dmarc.ietf.org>, > draft-ietf-bess-rfc7432...@ietf.org <draft-ietf-bess-rfc7432...@ietf.org> > *Cc: *bess-cha...@ietf.org <bess-cha...@ietf.org>, bess@ietf.org < > bess@ietf.org> > *Subject: *Re: Mail regarding draft-ietf-bess-rfc7432bis > > Hello Mathew, > > > > Just wondering if you received a response to your email, as I have not > seen any responses to either of our emails on the list. > > > > Thank you kindly. > > > > Best Regards, > > Menachem > > > > *From: *BESS <bess-boun...@ietf.org> on behalf of Matthew Bocci (Nokia) > <matthew.bocci=40nokia....@dmarc.ietf.org> > *Date: *Tuesday, 30 January 2024 at 17:42 > *To: *draft-ietf-bess-rfc7432...@ietf.org < > draft-ietf-bess-rfc7432...@ietf.org> > *Cc: *bess-cha...@ietf.org <bess-cha...@ietf.org>, bess@ietf.org < > bess@ietf.org> > *Subject: *Re: [bess] Mail regarding draft-ietf-bess-rfc7432bis > > CAUTION: External E-Mail - Use caution with links and attachments > > > > Hi Authors > > > > Resending this and including the WG. I believe this is a similar question > to the one posted by Menachem on RFC8214. > > > > Thanks in advance > > > > Matthew > > > > *From: *Matthew Bocci (Nokia) <matthew.bo...@nokia.com> > *Date: *Monday, 15 January 2024 at 12:40 > *To: *draft-ietf-bess-rfc7432...@ietf.org < > draft-ietf-bess-rfc7432...@ietf.org> > *Cc: *bess-cha...@ietf.org <bess-cha...@ietf.org> > *Subject: *Mail regarding draft-ietf-bess-rfc7432bis > > Hi Authors > > > > There is there following restriction (highlighted in yellow) on the use of > the control word in EVPN where the EL/ELI is used. I know this was > inherited from RFC7432, but do you know why this is the case (in particular > a SHOULD NOT)? > > > > The head end PE has no idea what hashing mechanism is actually used > downstream, regardless of whether the entropy label is inserted by it. The > entropy label is just there to provide additional flow information if the > downstream P router is load balancing based on the label stack, but it does > not in itself prevent the P router from scanning below the bottom of stack > and instead load balancing on the payload after checking the MPLS first > nibble. This also seems to be superseded by RFC8469 and all the discussion > over the years about making CW mandatory for MPLS-based services . It is > also worth noting that CW is not just to prevent aliasing between IP and > Ethernet traffic, but can be used to indicate OAM or other types of > maintenance packets. > > > > Can we just remove the text in yellow? > > > > Thanks > > > > Matthew > > > > > > In order to avoid frame misordering described above, the following > > network-wide rules are applied: > > > > * If a network uses deep packet inspection for its ECMP, then the > > the following rules for "Preferred PW MPLS Control Word" [RFC4385] > > apply: > > > > - It MUST be used with the value 0 (e.g., a 4-octet field with a > > value of zero) when sending unicast EVPN-encapsulated packets > > over an MP2P LSP. > > > > - It SHOULD NOT be used when sending EVPN-encapsulated packets > > over a P2MP or P2P RSVP-TE LSP. > > > > - It SHOULD be used with the value 0 when sending EVPN- > > encapsulated packets over a mLDP P2MP LSP. There can be > > scenarios where multiple links or tunnels can exist between two > > nodes and thus it is important to ensure that all packets for a > > given flows take the same link (or tunnel) between the two > > nodes. > > > > * If a network uses entropy labels per [RFC6790], then the control > > word SHOULD NOT be used. > > > > > _______________________________________________ > BESS mailing list > BESS@ietf.org > https://www.ietf.org/mailman/listinfo/bess >
_______________________________________________ BESS mailing list BESS@ietf.org https://www.ietf.org/mailman/listinfo/bess
