Re: [Anima] Self-Managed Networks

["Michael Behringer (mbehring)" <[e-mail]>]

Adding to Brian's response (which I agree with), and up-levelling the discussion a bit: 

Mehmet, I think there is a simple misunderstanding here. The ACP is a control plane for *device* management. A device can hold many contexts, as in NFV, but they are all managed in a single way. My feeling is we have a disconnect somewhere on this level, right? 

Michael

> -----Original Message-----
> From: Brian E Carpenter [mailto:[e-mail]]
> Sent: 16 October 2015 08:02
> To: Toy, Mehmet <[e-mail]>; Michael Behringer
> (mbehring) <[e-mail]>; [e-mail]
> Subject: Re: Self-Managed Networks
> 
> On 16/10/2015 16:36, Toy, Mehmet wrote:
> > Michael,
> >
> > Before I pick each of your statement and express my thoughts, I would like
> to ask more questions by picking your  statement " There is *one* ACP,
> there are *many* autonomic functions on top.".   Assume we  start building
> on this to see how far we can go, whether this holds for all cases or not, I
> don't know.
> >
> > If we have two or more ACPs, what could go wrong?
> 
> Why would a domain need more than one ACP? (Certainly there might be a
> transient situation after a temporary network partition where two ACPs
> would "meet" and then need to merge; indeed the ACP needs to be self-
> repairing in such cases.)
> 
> > What is it that you are trying to do here? What is the purpose behind this
> model?
> 
> Provide a secure L3 between all autonomic nodes that is strictly independent
> of the operational data plane. The purpose is to allow autonomic operations
> to work regardless of anything else.
> 
> >
> > Furthermore, I suggest you to think about ETSI NFV model,
> > "infrastructure" and "VNF" division of networks.  What is wrong in
> > using that kind of model here which is a reasonable breakdown,
> 
> We aren't reinventing NFV, as far as I can tell. If there is a recent technical
> overview of NFV, that would be interesting to see. But I don't believe that
> Autonomic Service Agents are at all the same thing as VNFs.
> 
> > instead of sticking with ACP  that is becoming very difficult to define?
> 
> Huh? draft-ietf-anima-autonomic-control-plane is still work in progress but it
> seems very clearly defined to me. In sketching out how GRASP will be
> implemented, I haven't hit any conceptual problems with the ACP, and its
> main API will just be socket calls. The only complication I found is that it will
> need to support the Advanced Socket API because of some special
> requirements for link-local multicasts. Otherwise, it's just another virtual
> (loopback) network interface.
> 
> Regards
>    Brian
> 
> >
> > Thanks
> > Mehmet
> >
> > -----Original Message-----
> > From: Michael Behringer (mbehring) [mailto:[e-mail]]
> > Sent: Thursday, October 15, 2015 4:57 AM
> > To: Toy, Mehmet; Brian E Carpenter; [e-mail]
> > Subject: RE: Self-Managed Networks
> >
> > Toy, thanks for raising those questions. Obviously, we're not doing a good
> job yet in describing what the ACP is, and that needs to be fixed. And
> obviously, we all need the same view to progress further. So this is a very
> important discussion, and I really welcome it.
> >
> > Before formalising better text, let me see whether we get agreement on
> the fundamental idea.
> >
> > In my head, there are two layers: The ACP, and on top of that the
> Autonomic Functions:
> >
> > * The ACP is the "tool kit". It comprises various "mechanics", such as
> negotiation, synchronisation, discovery of various sorts, messaging, etc.
> Those are all based on a common addressing and naming concept.
> >
> > * Autonomic Functions use that tool kit to do something clever. In other
> words, the true autonomic "intelligence" sits on that level.
> >
> > There is *one* ACP, there are *many* autonomic functions on top.
> >
> > One way to decide to which layer something belongs is to ask: "is this (1) a
> generic functionality which many functions require, or is this (2) one specific
> function?". If the answer is (1), it belongs into the ACP, if (2) it belongs into
> an autonomic function.
> >
> > So, in this light, my understanding (!) of fault management is that
> > this is an autonomic function, and would use common blocks of the
> > underlying ACP. Conversely, it would not offer services to other
> > autonomic functions on top. This is my way of thinking when I write
> > "this is an autonomic function".  And I'm not 100% certain I
> > understand what you're suggesting, so please chime in here! (And I
> > haven't read your draft fully yet, sorry)
> >
> > Look at
> > http://tools.ietf.org/id/draft-jiang-anima-prefix-management-01.txt
> > This draft describes "intelligence". In that case, a way to automatically
> manage address space. Sections 2.2 and 2.3 explain which parameters and
> information exchanges such a function would require. Sheng wrote this
> document to explain how an autonomic function would use a common ACP.
> >
> > Probably we should take some off-line time in Yokohama to discuss this in a
> small team?
> >
> > Michael
> >
> >
> >> -----Original Message-----
> >> From: Anima [mailto:[e-mail]] On Behalf Of Toy,
> Mehmet
> >> Sent: 15 October 2015 04:26
> >> To: Brian E Carpenter <[e-mail]>; Michael
> >> Behringer
> >> (mbehring) <[e-mail]>; [e-mail]
> >> Subject: Re: [Anima] Self-Managed Networks
> >>
> >> Michael and Brian,
> >> Per Toerless suggestion, I am including ANIMA group into the discussion.
> >>
> >> I re-read the "A Reference Model for Autonomic Networking" document
> >> and I am not clear about the definitions.
> >>
> >> a)  In the "A Reference Model for Autonomic Networking", ACP is
> >> defined as "The Autonomic Control Plane is the summary of all
> >> interactions of the Autonomic Networking Infrastructure with other
> nodes and services.".
> >>
> >> b) Brian, you write as " The ACP is common infrastructure for all
> >> autonomic functions.(The ACP needs to be self-repairing, of course.)
> >> The signaling protocol is also common infrastructure."
> >>
> >> Question: What is ACP? a or b or combination?
> >>
> >> c) Section 4 in the reference model document , "The Autonomic
> >> Networking Infrastructure provides a layer of common  functionality
> >> across an Autonomic Network.  It comprises "must implement" functions
> >> and services, as well as extensions."
> >> Question: What are the "must implement" functionalities?  How do you
> >> define "must implement" functionalities?
> >>
> >> Thanks
> >> Mehmet
> >> -----Original Message-----
> >> From: Brian E Carpenter [mailto:[e-mail]]
> >> Sent: Wednesday, October 14, 2015 5:25 PM
> >> To: Michael Behringer (mbehring); Toy, Mehmet
> >> Cc: '[e-mail]'; '[e-mail]'; 'anima-
> >> [e-mail]'
> >> Subject: Re: Self-Managed Networks
> >>
> >> I agree with Michael. The ACP is common infrastructure for all
> >> autonomic functions.
> >> (The ACP needs to be self-repairing, of course.) The signaling
> >> protocol is also common infrastructure.
> >>
> >>    Brian
> >> On 15/10/2015 05:43, Michael Behringer (mbehring) wrote:
> >>> I would argue they are part of an autonomic function, which runs on
> >>> top of
> >> the ACP.
> >>> There are really two different pieces here, and this is I think the
> >>> confusion
> >> here:
> >>>
> >>> -          The ACP is self-managing. It needs to do self-healing, and
> >> automatically adapt to new situations. But to me, this isn’t fault
> >> management or performance management as an operator understands
> it.
> >>>
> >>> -          The network has FM and PM function. Those could be (and should
> >> be, imo) autonomic functions. Those run on top of the ACP.
> >>> Bottom line: I’d like to keep the ACP itself as minimalistic and
> >>> simple as we
> >> possibly can. Functions like FM / PM belong into an autonomic function,
> IMO.
> >>> What do you think?
> >>> Michael
> >>> From: Toy, Mehmet [mailto:[e-mail]]
> >>> Sent: 14 October 2015 18:30
> >>> To: Michael Behringer (mbehring) <[e-mail]>
> >>> Cc: '[e-mail]' <[e-mail]>;
> >> '[e-mail]'
> >>> <[e-mail]>; '[e-mail]'
> >>> <[e-mail]>; '[e-mail]'
> >>> <[e-mail]>
> >>> Subject: RE: Self-Managed Networks
> >>>
> >>> Michael,
> >>> Instead of answering the question as Yes or No, let me give examples
> >>> to
> >> see what makes sense.
> >>> Let’s say in a data path, a router port is failed.  The router
> >>> generates an AIS
> >> (Alarm Indication Signal) and the receiving  end generates RDI
> >> (remote Defect Indicator).  Both messages are generated by the
> hardware, not by a
> >> software or ACP.   As a result of this failure,  there would be packet loss.
> The
> >> hardware counts these losses, an ACP does not.
> >>> For the FM and PM functions above, can we say they are part of an ACP?
> >>> Thanks
> >>> Mehmet
> >>> From: Michael Behringer (mbehring) [mailto:[e-mail]]
> >>> Sent: Wednesday, October 14, 2015 11:34 AM
> >>> To: Toy, Mehmet
> >>> Cc: '[e-mail]'; '[e-mail]'; 'anima-
> >> [e-mail]'; '[e-mail]'
> >>> Subject: RE: Self-Managed Networks
> >>>
> >>> Hi Toy,
> >>> To understand better: To me, fault management *uses* the functions
> >>> of
> >> the AN infrastructure. It uses the ACP to communicate, maybe GRASP
> >> for some signalling, might be influenced by Intent, etc. Right?  So
> >> to me, this is a logical component of an autonomic network that sits
> >> on top of the AN infrastructure.
> >>> Do we agree?
> >>> Michael
> >>>
> >>> From: Toy, Mehmet [mailto:[e-mail]]
> >>> Sent: 13 October 2015 23:46
> >>> To: Michael Behringer (mbehring)
> >>> <[e-mail]<mailto:[e-mail]>>
> >>> Cc: '[e-mail]'
> >>> <[e-mail]<mailto:[e-mail]>>;
> >>> '[e-mail]'
> >>> <[e-mail]<mailto:[e-mail]>>;
> >>> '[e-mail]'
> >>> <[e-mail]<mailto:[e-mail]>>;
> >>> '[e-mail]'
> >>> <[e-mail]<mailto:[e-mail]>>
> >>> Subject: RE: Self-Managed Networks
> >>>
> >>> Michael,
> >>> Appreciate the reply.
> >>> FM is part of data plane and control plane (i.e. ANI in your diagram).
> >>> My plan is to add a short paragraph for now either to section 2 to
> >>> expand
> >> the description  of ANI or to section 4 to add a sub-section for
> >> Fault Management.
> >>>
> >>> It is also possible too add a Performance Management section to
> >>> describe
> >> what types of measurements and where and how are used.  Although
> >> there is a control feedback related measurement in the document, I
> >> don’t know if it is adequate.
> >>>
> >>> Regards,
> >>> Mehmet
> >>> From: Michael Behringer (mbehring) [mailto:[e-mail]]
> >>> Sent: Tuesday, October 13, 2015 12:27 PM
> >>> To: Toy, Mehmet
> >>> Cc: '[e-mail]'; '[e-mail]'; 'anima-
> >> [e-mail]'; '[e-mail]'
> >>> Subject: RE: Self-Managed Networks
> >>>
> >>> Sorry for the delay, it’s very busy at the moment here.
> >>> To me, fault management refers generally to faults on the data
> >>> plane, ie for
> >> user traffic. I see that happening at some point as an autonomic
> >> function (or several, for different aspects). Would you agree? Or do
> >> you see that as a function inside the AN infrastructure?
> >>> So my feeling is that function would reside on top of the
> >>> infrastructure that
> >> we’re currently defining. So, please have a look whether your
> >> thoughts can be described as an autonomic function. I think they probably
> can.
> >>> Then I suggest we do the same that we’re planning to do with the NMS
> >> section, the model discussion, etc: Have a short paragraph describe
> >> the overall topic briefly, and point to an external doc for now,
> >> i.e., probably your draft.
> >>> If you agree, can you suggest where in the reference model you would
> >>> add
> >> a short paragraph about fault management, and I suppose we’d point to
> >> your draft, right?
> >>> Michael
> >>>
> >>>
> >>> From: Toy, Mehmet [mailto:[e-mail]]
> >>> Sent: 13 October 2015 03:53
> >>> To: Michael Behringer (mbehring)
> >>> <[e-mail]<mailto:[e-mail]>>
> >>> Cc: '[e-mail]'
> >>> <[e-mail]<mailto:[e-mail]>>;
> >>> '[e-mail]'
> >>> <[e-mail]<mailto:[e-mail]>>;
> >>> '[e-mail]'
> >>> <[e-mail]<mailto:[e-mail]>>;
> >>> '[e-mail]'
> >>> <[e-mail]<mailto:[e-mail]>>
> >>> Subject: RE: Self-Managed Networks
> >>>
> >>> Mike,
> >>> I am waiting for your response.
> >>> Thanks
> >>> Mehmet
> >>>
> >>> From: Toy, Mehmet
> >>> Sent: Friday, October 09, 2015 8:13 PM
> >>> To: '[e-mail]'
> >>> Cc: '[e-mail]'; '[e-mail]'; 'anima-
> >> [e-mail]'; '[e-mail]'
> >>> Subject: RE: Self-Managed Networks
> >>>
> >>> Mike,
> >>> I can send you some text to include in section 2 and 4 of  “A
> >>> Reference
> >> Model for Autonomic Networking,   draft-behringer-anima-reference-
> >> model-03”,  per Sheng’s suggestion.
> >>> Should I just do that?
> >>> Thanks
> >>> Mehmet
> >>>
> >>>
> >>> From: Toy, Mehmet
> >>> Sent: Thursday, October 08, 2015 7:53 AM
> >>> To: '[e-mail]'; '[e-mail]';
> >> '[e-mail]'; '[e-mail]'
> >>> Cc: '[e-mail]'
> >>> Subject: Re: Self-Managed Networks
> >>>
> >>> Sheng,
> >>> Appreciate a quick response.
> >>> I will work on your suggestion.
> >>> Mehmet
> >>>
> >>>
> >>> From: Sheng Jiang [mailto:[e-mail]]
> >>> Sent: Thursday, October 08, 2015 06:08 AM Eastern Standard Time
> >>> To: Toy, Mehmet;
> >>> [e-mail]<mailto:[e-mail]>
> >>> <[e-mail]<mailto:[e-mail]>>;
> >>> [e-mail]<mailto:[e-mail]>
> >>> <[e-mail]<mailto:[e-mail]>>;
> >>> [e-mail]<mailto:[e-mail]>
> >>> <[e-mail]<mailto:[e-mail]>>
> >>> Cc: Romascanu, Dan (Dan)
> >>> ([e-mail]<mailto:[e-mail]>)
> >>> <[e-mail]<mailto:[e-mail]>>
> >>> Subject: RE: Self-Managed Networks
> >>>
> >>> Hi, Toy,
> >>> First of all, for my understanding, your work is in the scope of the
> >>> WG
> >> charter. However, we do not have work item or milestone for it. It
> >> looks like an upper-layer autonomic service agent for me. In our
> >> plan, autonomic service agents are mainly for the next period, which
> >> is after re-charter (this is the same with your suggestion of
> >> modifying the charter, but it cannot happen until we deliver the
> >> current milestones). For now, the best may be try to add some
> >> description, maybe mainly abstracted functionality, into the reference
> model document.
> >>> Best regards,
> >>> Sheng
> >>>
> >>> From: Toy, Mehmet [mailto:[e-mail]]
> >>> Sent: Tuesday, October 06, 2015 1:14 AM
> >>> To: [e-mail]<mailto:[e-mail]>;
> >>> [e-mail]<mailto:[e-mail]>;
> >>> [e-mail]<mailto:[e-mail]>
> >>> Cc: Romascanu, Dan (Dan)
> >>> ([e-mail]<mailto:[e-mail]>)
> >>> Subject: Self-Managed Networks
> >>>
> >>> Dear All:
> >>> I couldn’t attend the Prague meeting, but luckily Dan was able to
> >>> present
> >> my slides on “Self-Managed Networks with Fault Management
> Hierarchy”.
> >> The feedback was to position the work in the ANIMA WG scope and
> >> framework.
> >>>
> >>> ANIMA charter in “M. Behringer, et. al., A Reference Model for
> >>> Autonomic
> >> Networking
> >>> draft-behringer-anima-reference-model-03” refers to “self-healing”.
> >> RFC7575,  “M. Behringer, et al.,   Autonomic Networking: Definitions and
> >> Design Goals”,  refers to “self-management”. However, both documents
> >> do not  articulate fault management aspect of the self-management.
> >> It is possible to interpret the fault management aspect of autonomic
> >> networks as part of “self-healing” and therefore as part of the ANIMA
> >> charter.  In that case, the “Architectural Framework for Self-Managed
> >> Networks with Fault Management Hierarchy,
> >> draft-mtoy-anima-self-faultmang-framework-
> >> 00.txt” contribution can target to fill that gap.  The control plane
> >> aspect of self-healing is addressed by “M. Behringer, et al., An
> >> Autonomic Control Plane,
> >> draft-behringer-anima-autonomic-control-plane-03”.  I believe these
> >> contributions are complementary to each other. I can try to address that
> in the contribution.
> >>>
> >>> Please let me know if you agree with me. If not, I suggest to modify
> >>> the
> >> charter since without covering fault management aspect of the
> >> autonomic networks, the concept of autonomic network will be
> incomplete.
> >>>
> >>> Regards,
> >>> Mehmet
> >>>
> >>>
> >>
> >>
> >> _______________________________________________
> >> Anima mailing list
> >> [e-mail]
> >> https://www.ietf.org/mailman/listinfo/anima