Re: [Rtg-ooam-dt] [sfc] Identifying OAM in NSH

["Carlos Pignataro (cpignata)" <[e-mail]>]

Dave,

[Apologies for the delay in responding — I was on vacation]

I appreciate the healthy tension between over- and under-specifying. I do not think however that the behavior is undefined in this particular case.

To me, the specification is: “when O=1, then the packet is an OAM packet processed by an OAM module” with the corollaries regarding the OAM handing: (with the OAM module processing defined in a separate place) and (if the OAM module is NULL (does not exist), then the packet goes to the bit-bucket)

O=1 does not mean an OAM header follows. It means it’s an OAM packet. And OAM is not a protocol type, it’s a set of functionalities realized by various protocols potentially.

Thanks!

— Carlos.

On Jul 24, 2016, at 2:25 PM, Dave Dolson <[e-mail]<mailto:[e-mail]>> wrote:

Carlos,
I don't find it satisfying  to define a protocol which says if O=1 the behavior is undefined. (Same outcome as when version!=0)

But, while the NSH draft is mute on the topic, it is fair to speculate on various  schemes.

‎Trailing an OAM header is an alternative to putting piggyback OAM in the metadata. I'm not sure where I heard it mentioned. Perhaps it was in the context of unified Overlay OAM. Logically if O=1, there must be an OAM header, right?

> If O=1 and the next protocol is IPv4, I’d expect an IPv4 packet
> encapsulating the OAM (either UDP-> BFD, or ICMP, for example).
Do you mean check to see if the embedded IPv4 packet is addressed to
the SFF itself? I.e., send the packet up the local ipv4 stack?




From: Carlos Pignataro (cpignata)
Sent: Saturday, July 23, 2016 7:29 PM
To: Dave Dolson
Cc: Gregory Mirsky; [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [sfc] [Rtg-ooam-dt] Identifying OAM in NSH


Hi, Dave,

On Jul 23, 2016, at 6:38 PM, Dave Dolson <[e-mail]<mailto:[e-mail]>> wrote:

Carlos,
I think some of what is being said in emails needs to be clarified in the document.
It still isn't crystal clear to me.


I agree that it needs to be clarified, with diagrams, and specs. However, when you say “in the document”, I do not think those details belong in the NSH document itself. The NSH, in my humble opinion, needs to say how to identify an OAM packet.

If O=1, and next protocol is IPv4, can there be an OAM header piggy-backed with the IPv4? If so, before or after?
Some drafts imply this is possible but nothing says how.

If O=1 and the next protocol is IPv4, I’d expect an IPv4 packet encapsulating the OAM (either UDP-> BFD, or ICMP, for example). With “OAM header”, do you mean the OOAM DT header? What would be the use for those extra bytes in this case?

Thanks,

— Carlos.


-Dave


From: Carlos Pignataro (cpignata)
Sent: Saturday, July 23, 2016 12:25 PM
To: Gregory Mirsky
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [sfc] [Rtg-ooam-dt] Identifying OAM in NSH


Hi, Greg,

On Jul 22, 2016, at 10:24 AM, Gregory Mirsky <[e-mail]<mailto:[e-mail]>> wrote:

Hi Carlos,
thank you for sharing your comments. If I understand correctly, you suggest to expose protocol types of OAM as Next Protocol rather than to use single Active OAM protocol type and use OOAM Header to demux OOAM type. Then, the Next Protocol registry will have to allocate values for single-hop BFD, multi-hop BFD, multipoint BFD, S-BFD, Echo Request/Reply, AIS protocol, Active Performance Measurement protocol, and I’ve only listed some of AOM protocols that may be used to operate, administer and maintain SFP.

Yes, the protocol field ought to register the OAM protocols that will be used and implemented and deployed — of course not all potential variations and permutations of possible OAMs (what is AIS protocol?)

Additionally, you’ve suggested that only O-bit value to be used to determine whether packet should be processed as OAM or data packet. Hence should I expect that O-bit is set for BFD, AIS, and Echo Request/Reply payload and should not be set if the Next Protocol is neither of protocols listed above? Should such situation, i.e. Next Protocol value is MPLS and O-bit set to 0x1, should be viewed as error and the packet dropped? Or you propose that the Next Protocol takes precedence and the packet treated as data? Or packet viewed as OAM and passed to the local OAM entity? Or how to interpret situation when O-bit is cleared and the Next Protocol value is one of OAM protocols? This is the situation that, in my view, is ambiguous and under-specified in the current NSH document. My proposal is an attempt to make relationship between OAM and data packets more deterministic.

Answering all those questions (which are really slight permutations of the same question) in one shot: to me, O=0 is a data packet and O=1 is an OAM packet. If the data processing does not have a handler for the protocol in the PID, or it’s an undefined, drops to the bit bucket. Equally, if the OAM handler does not support the protocol ID carried as OAM, puff. IP can carry data or OAM for example by the way.

It does not get any simpler and unambiguous than that. What’s the advantage of moving the OAM PID further inside?

And I do not believe there’s underspecification in NSH -> O=1, OAM treatment, not detailed here.

Thanks,

— Carlos.


                Regards,
                                Greg

From: Carlos Pignataro (cpignata) [mailto:[e-mail]]
Sent: Friday, July 22, 2016 10:10 AM
To: Gregory Mirsky <[e-mail]<mailto:[e-mail]>>
Cc: [e-mail]<mailto:[e-mail]>; [e-mail]<mailto:[e-mail]>
Subject: Re: [Rtg-ooam-dt] Identifying OAM in NSH

Greg,

Please find some comments inline.
Thumb typed by Carlos Pignataro.
Excuze typofraphicak errows

On Jul 21, 2016, at 09:01, Gregory Mirsky <[e-mail]<mailto:[e-mail]>> wrote:
Dear All,
we had very good discussion on OAM this week. We’ve touched on Active, Passive and Something-in-between OAM. But can NSH indicate which OAM type, if any, associated with a packet? I think that the current version of draft-ietf-sfc-nsh does not allow that and may be ambiguous in some cases. I propose change to interpretation and applicability description of the O(AM) flag and allocation of the new protocol type to be used in the Next Protocol field.


Active, passive and something-in-between are not OAM protocol types but rather they are measuring methods. Active OAM can include a plurality of OAM protocols, including BFD, S-BFD, something-over-ip, etc.

I also believe that the current OAM text in NSH is not ambiguous and allows enough processing of the header to understand something is OAM, without going the specifics of an OAM handler.

Therefore, I oppose this change.


RFC 7799 defines Active OAM as following:
An Active Metric or Method depends on a dedicated measurement packet stream and observations of the stream.
Thus, regardless of encapsulation used by OAM, it is the packet constructed solely for monitoring, measuring network’s metric and should not leave given domain. And, I believe, such packets should be identified by the protocol type of their own.

Seems you are advocating for a single "OAM" protocol type for OAM packets. The functionality of identifying something as OAM is in the O-bit, so no need to waste bits in duplication.

Then, at some point, you have to differentiate if an OAM is, say, IP or "raw BFD" or something else. That's what the Protocol field is for. I do not see a need to add an indirection here to then have to have a protocol field inside the OAM.


OAM which behaves as much as Passive OAM or Something-in-between, e.g. OAM appended to data packet either at the domain’s edge or on-the-fly, identified by the protocol type of the data packet carried in NSH.

That's correct, with the O bit cleared; it's a data packet and not an OAM one.


Below are changes I’d like to propose:
Section 3.2 O-bit:
OLD
   O bit: when set to 0x1 indicates that this packet is an Operations,
   Administration and Maintenance (OAM) message.  The receiving SFF and
   SFs nodes MUST examine the payload and take appropriate action (e.g.
   return status information).  OAM message specifics and handling
   details are outside the scope of this document.
END
NEW
O bit: when set to 0x1 indicates that data packet identified by the Next
Protocol type has OAM metadata appended.

No. If it is a data packet it does not have the O bit set (to 1 or to whichever other possible value for the bit :-)

The goal is that looking at a single but it can be understood if it is a data packet (which can be used, marked, etc. to be used for OAM purposes, or not).

We do not want OAM direct exception processing for data packets.


Definition of the format(s)
used by OAM metadata is outside the scope of this document.
END

At the end of Section 3.2:
OLD
   This draft defines the following Next Protocol values:

   0x1 : IPv4
   0x2 : IPv6
   0x3 : Ethernet
   0x4: NSH
   0x5: MPLS
   0x6-0xFD: Unassigned
   0xFE-0xFF: Experimental
END
NEW
   This draft defines the following Next Protocol values:

   0x1 : IPv4
   0x2 : IPv6
   0x3 : Ethernet
   0x4: NSH
   0x5: MPLS
   0x6: Active OAM

As per above I do not believe there's a single OAM protocol.


   0x7-0xFD: Unassigned
   0xFE-0xFF: Experimental
END

And, consequently, section 13.2.5 in IANA Considerations section will request allocation of value 0x6 to be assigned to Active OAM protocols.

Greatly appreciate your consideration and comments.


My €0.02.

Thanks,

Carlos.


                Regards,
                                Greg

_______________________________________________
Rtg-ooam-dt mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/rtg-ooam-dt