Re: [spring] Request comments/feedback on https://datatracker.ietf.org/doc/draft-zzhang-spring-microtap-segment/01/

["Jeffrey (Zhaohui) Zhang" <[e-mail]>]

Hi Eduard,

Please see zzh2> below.



Juniper Business Use Only
From: Eduard Metz <[e-mail]>
Sent: Sunday, March 3, 2024 7:14 AM
To: Jeffrey (Zhaohui) Zhang <[e-mail]>
Cc: Jeff Tantsura <[e-mail]>; Ryan Hoffman <ryan.hoffman=[e-mail]>; [e-mail]; [e-mail]
Subject: Re: [spring] Request comments/feedback on https://datatracker.ietf.org/doc/draft-zzhang-spring-microtap-segment/01/

[External Email. Be cautious of content]


Thanks, see inline / below.

/Eduard



Few comments after first read:
- For SRv6 the procedure may be slightly different, ie steer traffic via MicroTAP capable node or have MicroTAP as integrated capability of "default" forwarding (of capable nodes) and indicate the parameter - this is the approach in the current draft if I understand correctly.

Zzh> The microtap segment belongs to the node connected to the monitor, which is typically not in the path of most traffic. When a capable node in the normal traffic path encounters a microtap SID (which is not advertised by that node), it makes a copy and send the copy to the owner of the microtap SID (while continue to forward the original copy after removing the microtap SID).
Zzh> Therefore, it is not an integrated capability of “default” forwarding (of capable nodes).


[EM] Makling a copy based on the microtap SID (instead of simply forwarding it) is what I referred to as an integrated capabiltiy of the basic SR forwarding. In SRv6 alternatively one could add a SID that points (explicitly) to the microtap (replication) function (with the reference to the microtap SID as the argument) , instead of triggering a different behaviour based on the value of a SID.

Zzh2> The microtap SID from the node that connects to the monitor does explicitly point to the microtap function on each tapping node. That is analogues to that a prefix SID from a node has a function on each node as “forward to the prefix”.
Zzh2> If I understand you correctly, let’s say if the normal explicit path is (S1, S2, S3, S4) where each Sx is a node SID for a node Nx, you’d use S1, S2’, S3, S4 when you want N2 to tap it, where S2’ is N2’s microtap SID for a particular monitor.
Zzh2> This means that each tapping-capable node needs to advertise an additional SID. That may not be desired, but we can discuss if that should be a preferred solution or an alternative.
Zzh2> For comparison, the method in the draft is that an additional microtap SID is inserted after S2: (S1, S2, Tm, S3, S4). Tm is the microtap SID advertised by the node connecting to the monitor.
- Section 2.3 describes that if a MicroTAP SID becomes the active on the a node not supporting the MicroTAP capability, the packet would be dropped. I wondered if this is correct? Wouldnt the packet be forwarded to the "monitor" node? This breaks the communication effectively, but not a drop at the node not supported MicroTAP.

Zzh> A node not supporting MicroTAP will not advertise its capability or install the forwarding state for MicroTAP SIDs (advertised by the nodes connected to the monitors).
Zzh> As a result, other nodes SHOULD NOT place a MicroTap SID after the node/adj SID for the incapable node. In the unlikely case if that happened, in the case of MPLS the packet will simply be dropped (there is no corresponding state). In the case of SRv6, there might not be a corresponding IPv6 route either and traffic will also be dropped. But if there is a less specific route covering that MicroTap SID, then it will be forwarded accordingly. We will add that clarification.

[EM] Ah, this is due to the SID type? Some clarification would be useful indeed.
Zzh2> Not due to the SID type, but due to the fact that the SID is simply an IPv6 address, so a less specific route may be matched.
Zzh2> Thanks!
Zzh2> Jeffrey


- In general, or least for intercept, one would be interested in both directions of a traffic stream (e.g to / from a specific IP). To address this, the MicroTAP SID would need to inserted on all relevant ingresses. And the monitor may receive packets from different MicroTAP capable nodes. This may have implications for the use of IOM header (e.g. to avoid duplicate sequence ids)

Zzh> A monitor is going to receive tapped packets from all over the places (it all depends on which packets carry the MicroTap SID and where in the SID list), but unless a MicroTap SID is repeated (in different places of the SID list) in the packet, the monitor will only receive one tapped copy for a particular packet. I also imagine that an ingress is likely coordinating with the monitor when it places the MicroTap SID, even though that’s outside the scope of this draft.
Zzh> Can you explain the implications for the use of IOM header?

[EM] I wondered whether the sequence ID would be sufficient or additional info is needed, e.g. to identify a tapped traffic stream and establish if all tapped packets have been received at the monitor. At the source one may want to add additional information anyway to identify a tapped traffic stream, e.g. to combine and/or demultiplex packets at the monitor.




Zzh> Thanks.
Zzh> Jeffrey

cheers,
  Eduard


On Wed, Feb 28, 2024 at 4:52 AM Jeff Tantsura <[e-mail]<mailto:[e-mail]>> wrote:
Seems like a very useful feature indeed.

Cheers,
Jeff

On Feb 27, 2024, at 07:15, Ryan Hoffman <ryan.hoffman=[e-mail]<mailto:[e-mail]>> wrote:

TELUS intends to deploy this microTap segment feature once available in vendor NOS after thorough testing in our lab.  We'd expedite TELUS testing and deployment when available from vendors, as this is a much needed feature in our network.

Thanks,
Ryan Hoffman

On Wed, Apr 5, 2023 at 3:28 PM Jeffrey (Zhaohui) Zhang <[e-mail]<mailto:[e-mail]>> wrote:
Hi,

The authors of this draft would like to get your feedback on this draft.

   This document specifies a microTap segment that can be used to
   instruct a transit node to make a copy of a segment-routed packet and
   deliver it to a specified node for the purpose of network monitoring,
   trouble shooting, or lawful intercept.

Due to the limit of Spring WG session time we have not been able to present it but we submitted slides before: https://datatracker.ietf.org/meeting/115/materials/slides-115-spring-slides-115-spring-microtap-segment-00<https://urldefense.com/v3/__https:/datatracker.ietf.org/meeting/115/materials/slides-115-spring-slides-115-spring-microtap-segment-00__;!!NEt6yMaO-gk!EKhg-4oZEfTFYJNmgp8IGr1V5a4BR45VWyFGE1yjXKX5wyy_b1J1I5V1a1TwceJBWn4B_S11vdYmbIo$>.

Thanks.
Jeffrey

Juniper Business Use Only
_______________________________________________
spring mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/spring<https://urldefense.com/v3/__https:/www.ietf.org/mailman/listinfo/spring__;!!NEt6yMaO-gk!EKhg-4oZEfTFYJNmgp8IGr1V5a4BR45VWyFGE1yjXKX5wyy_b1J1I5V1a1TwceJBWn4B_S11CmS-d-Q$>
_______________________________________________
spring mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/spring<https://urldefense.com/v3/__https:/www.ietf.org/mailman/listinfo/spring__;!!NEt6yMaO-gk!EKhg-4oZEfTFYJNmgp8IGr1V5a4BR45VWyFGE1yjXKX5wyy_b1J1I5V1a1TwceJBWn4B_S11CmS-d-Q$>