Re: [Teas] FW: The word "transport"

[Adrian Farrel <[e-mail]>]

> This is an obvious segue, quite possibly into an equally obvious rat-hole.

 

So, I should resist the lure of your rat-hole?

But it smells of such wonderful cheese.

 

> Nobody is (as far as I am aware) arguing that we cannot use existing

> IETF technology to “slice” IP networks using existing technology, for

> carrying arbitrary data (including control and management data) in

> packetized form.

 

Indeed.

And the point I was making is that we can also use IETF technologies to slice networks that carry arbitrary data that t not in a packetized form (as far as we are aware).

 

> Some folks have stated that current technology may not be enough.

> That remains to be seen.  We need to get a much better understanding

> of exactly what is required in order to make a decision on this, both in

> the short and long term.

 

Yes. I would hope we can focus on the service characteristics before working on protocol solutions.

 

> Another group (probably containing a subset of the same people) 

> are arguing that we will likely need a high-level model for creating,

> maintaining, monitoring, modifying, and removing a “service” as

> defined at that level.  The intent is that this work would take place

> at a level of abstraction where it is independent of specific

> technologies that may be used within the service, to deliver a

> requested service at that same level of abstraction. 

 

Sounds good to me. But then I have a history of working with “service level YANG models.”

 

> Actual work to define this model was not an objective (at least

> initially – and I believe – so far) for the design team, although

> some individual work has begun.

 

Sure. The DT’s scope and charter are quite clear. But, of course, anyone is free to work on anything and post any draft.

 

> We call this “service abstraction” a “transport slice service.” 

 

Well, why not?

Except, of course, for thinking that we should simply call this a “network slice service” because, after all, we are slicing the network.

 

> Implicit in this argument is a requirement for a function or

> component that can accept the information in this model

> and translate it into specific control and management

> instructions required to make the underlying network

> behave in the ways requested for the services to be

> provided.

> 

> We call this function/component a “transport slice controller

> (TSC).”  

 

Is that an abstract function? I like abstract functions. Architectures should have lots of them.

 

Of course, you can realise an abstract function by making an implementation. 

 

> As part of this controller (or control function) an interface is a

> need to define how these abstract service requests are handled.  

> 

> We call this interface the TSC’s Northbound Interface (NBI).

> 

> There is no current intent (at least on the part of most of the

> design team) to standardize this entity or it’s interactions with

> other entities in any network application, other than via this NBI.

 

So, this is both an abstract interface and a protocol specification. That’s OK.

And if other protocols are specified to satisfy the abstract interface that would also be fine (although we tend to like to have just one solution for any one problem).

 

> Other terminology is included to allow for logical descriptions of

> the required interaction between an arbitrary implementation

> of a TSC and the underlying network (which it is used to control)

> in providing any arbitrary set of transport slice services.

 

s/which it is used to control/which it controls/

 

> Example of this include the Southbound Interface (SBI), and

> the Network Controller – which may or may not exist in 

> every instance, especially if the TSC itself is not to be

> standardized, or if specific network controllers are not

> needed in some instances.

 

The TSC would never be standardised in the IETF, I think. We don’t standardize components beyond architectural work. 

 

> This is really not the sort of rocket surgery that warrants

> this much confusion, uncertainty and doubt.

 

Nevertheless, I am always grateful to have my doubts assuaged, my uncertainty clarified, my confusion swept away, and my rockets surged.

 

But wait! How did we get here?

 

Best,

Adrian

 

From: Adrian Farrel <[e-mail] <mailto:[e-mail]> > 
Sent: Friday, May 1, 2020 12:13 PM
To: Eric Gray <[e-mail] <mailto:[e-mail]> >; 'Rokui, Reza (Nokia - CA/Ottawa)' <[e-mail] <mailto:[e-mail]> >; 'John E Drake' <jdrake=[e-mail] <mailto:jdrake=[e-mail]> >; [e-mail] <mailto:[e-mail]> 
Subject: RE: [Teas] FW: The word "transport"
Importance: High

 

Eric,

 

I appreciate your attempt here, but I see no reason why we would not slice a sub-IP network utilising the technology that we have developed in the IETF for control and management of sub-IP networks.

 

In fact, maybe we should not try to use this thread to define the purpose and scope of the IETF 😊 The creation of CCAMP was an old argument and results in the IP-based control of non-packet networks being in scope.

 

Best,

Adrian

 

From: Eric Gray <[e-mail] <mailto:[e-mail]> > 
Sent: 01 May 2020 16:46
To: Rokui, Reza (Nokia - CA/Ottawa) <[e-mail] <mailto:[e-mail]> >; [e-mail] <mailto:[e-mail]> ; 'John E Drake' <jdrake=[e-mail] <mailto:jdrake=[e-mail]> >; [e-mail] <mailto:[e-mail]> ; [e-mail] <mailto:[e-mail]> 
Subject: RE: [Teas] FW: The word "transport"

 

Possibly it would be better to clarify this in terms of what the IETF does deal with?

 

The IETF deals with “packetized” (more specifically, where packetization uses IETF defined packet formats) transport of otherwise arbitrary data.

 

Here you can replace transport with “carrying” – but this is essentially the standard English meaning of transport.

 

From: Teas-ns-dt <[e-mail] <mailto:[e-mail]> > On Behalf Of Rokui, Reza (Nokia - CA/Ottawa)
Sent: Friday, May 1, 2020 11:38 AM
To: [e-mail] <mailto:[e-mail]> ; 'John E Drake' <jdrake=[e-mail] <mailto:jdrake=[e-mail]> >; [e-mail] <mailto:[e-mail]> ; [e-mail] <mailto:[e-mail]> 
Cc: Rokui, Reza (Nokia - CA/Ottawa) <[e-mail] <mailto:[e-mail]> >
Subject: Re: [Teas-ns-dt] [Teas] FW: The word "transport"

 

Adrian,

 

>>>>>> I don’t understand “we (IETF) do not deal with network slice”.

 

Please refer to Figure 4 of the draft.

In summary, an E2E network slice is the logical network from End user X to End use Y and comprises of:

*	One or more Transport slices (TS)
*	One or more Other Slices (OS)

 

Just as an example for “Other Slices”, in 5G these slices are RAN slices and 5G Core Slices.

Referring  this Figure, it is clear that IETF deals with Transport slice which is part of an E2E network slice and “Other slices” are outside scope of IETF.

Please make a clear distinction between Transport Slices and e2e network slices. They are different.

IMHO, any reference to “Network Slice” in IETF drafts means “Transport slice” . 

 

>>>>>>.  Why do we have a “TEAS Network Slicing Design Team”?

Good question. The correct name can be “TEAS Transport Slicing Design Team”

 



 

Reza 

 

 

From: Adrian Farrel <[e-mail] <mailto:[e-mail]> >
Organization: Old Dog Consulting
Reply-To: "[e-mail] <mailto:[e-mail]> " <[e-mail] <mailto:[e-mail]> >
Date: Friday, May 1, 2020 at 10:30 AM
To: Reza Rokui <[e-mail] <mailto:[e-mail]> >, 'John E Drake' <jdrake=[e-mail] <mailto:jdrake=[e-mail]> >, "[e-mail] <mailto:[e-mail]> " <[e-mail] <mailto:[e-mail]> >, "[e-mail] <mailto:[e-mail]> " <[e-mail] <mailto:[e-mail]> >
Subject: RE: [Teas] FW: The word "transport"

 

Hi Reza,

 

I don’t understand “we (IETF) do not deal with network slice”.

Why do we have a “TEAS Network Slicing Design Team”?

 

Cheers,

Adrian

 

From: Rokui, Reza (Nokia - CA/Ottawa) <[e-mail] <mailto:[e-mail]> > 
Sent: 01 May 2020 15:13
To: [e-mail] <mailto:[e-mail]> ; 'John E Drake' <jdrake=[e-mail] <mailto:jdrake=[e-mail]> >; [e-mail] <mailto:[e-mail]> ; [e-mail] <mailto:[e-mail]> 
Cc: Rokui, Reza (Nokia - CA/Ottawa) <[e-mail] <mailto:[e-mail]> >
Subject: Re: [Teas] FW: The word "transport"

 

John and all,

 

This is where I disagree with term “Underlay Network Slice”. It brings lots of other questions since we (IETF) do not deal with network slice as clearly described in our draft.

 

As I send in another email thread, as per Adrian’s suggestion, we define the term “Transport” as follows and would like to keep the term “Transport Slice”

 

*	Definition of term Transport: This term embraces various technologies such as IP, Optical, Ethernet, TDM, etc  that carry packets and traffic and might span multiple administrative domains.

 

Please provide your suggestion to refine this definition.

 

Cheers,

Reza

 

From: Adrian Farrel <[e-mail] <mailto:[e-mail]> >
Organization: Old Dog Consulting
Reply-To: "[e-mail] <mailto:[e-mail]> " <[e-mail] <mailto:[e-mail]> >
Date: Friday, May 1, 2020 at 5:23 AM
To: 'John E Drake' <jdrake=[e-mail] <mailto:jdrake=[e-mail]> >, Reza Rokui <[e-mail] <mailto:[e-mail]> >
Cc: "[e-mail] <mailto:[e-mail]> " <[e-mail] <mailto:[e-mail]> >
Subject: RE: [Teas] FW: The word "transport"

 

Yeah, ‘underlay’ seems good because it is a relative term and usefully recursive.

 

A

 

From: Teas <[e-mail] <mailto:[e-mail]> > On Behalf Of John E Drake
Sent: 30 April 2020 20:59
To: [e-mail] <mailto:[e-mail]> ; 'Rokui, Reza (Nokia - CA/Ottawa)' <[e-mail] <mailto:[e-mail]> >
Cc: [e-mail] <mailto:[e-mail]> 
Subject: Re: [Teas] FW: The word "transport"

 

Hi,

 

How about ‘underlay network slice’?  It’s describing exactly what we are doing and is congruent with the Enhanced VPN draft.

 

Yours Irrespectively,

 

John

 

 

Juniper Business Use Only

From: Teas <[e-mail] <mailto:[e-mail]> > On Behalf Of Adrian Farrel
Sent: Thursday, April 30, 2020 3:32 PM
To: 'Rokui, Reza (Nokia - CA/Ottawa)' <[e-mail] <mailto:[e-mail]> >
Cc: [e-mail] <mailto:[e-mail]> 
Subject: [Teas] FW: The word "transport"

 

[External Email. Be cautious of content]

 

Reza,

 

Thanks for forwarding this email.

 

I appreciate and understand the way that 3GPP use the term “transport”. It makes a lot of sense in the 3GPP context: they are talking about connectivity in their realm, and anything that provides that connectivity (such as an IP network) counts as “transport”.

 

But we are not the 3GPP and our terminology needs to be consistent. As I noted, we already have multiple meanings of transport:

*	The transport layer (traditional from the OSI model) that includes TCP, UDP, etc.
*	Transport networks (a term also used in the ITU-T) that embraces Ethernet, TDM, and optical technologies that carry packets for us. This term is most often seen in CCAMP, TEAS, and PCE.
*	MPLS-TP (possibly deriving from the ITU-T’s T-MPLS) that refers to the use of MPLS as a technology to build transport networks as described in the previous bullet
*	Transport as a verb (such as in HTTP) meaning simply ‘to carry’

In the ACTN work, we moved from ‘transport’ to ‘TE’ recognising that ACTN was applicable to any network type where paths could be computed and imposed, and resource partitioned and reserved.

 

Now, you say Transport networks (which are technology specific) are used to realized “Transport Slices” which would appear to imply that IP cannot be used to realise a transport slice. I don’t know if that is your intention.

 

You also say They [3GPP] did not define anything related to Transport but you say that immediately under a figure you have copied from 3GPP TS 28.530 that clearly shows transport slices, so that leaves me more than a little confused.

 

Additionally, you say…

VPN is one of the  solutions to realize the transport slices in IP networks.

…which is fine by me since I am not (here) talking about solutions but services. Actually, I think you would do well to distinguish between VPN as a service and the technologies used to realise a VPN.

 

Then…

However, The idea of Transport Slice is to allow a high-level system (or consumer or an Orchestrator) to ask for a various connections (i.e. Transport slices)

… This is pretty meaningless the way you have worded it. You have said “the idea of a transport slice is to all a request for transport slices...

across  IP, Optics, PON, Microwave or any combination of these networks. We shall not limit ourselves to IP VPN.

…Yes, indeed. I wonder where the idea of such a limitation came from…

Note that the definition of the Transport slice is technology agnostic.

…I know. It comes from draft-ietf-teas-enhanced-vpn. That originated in draft-king-teas-applicability-actn-slicing. I believe I drafted it…

[snip]

In summary, since the connectivity between various endpoints are across transport network (i.e. IP, Optics, PON, Microware etc.), it is logical to assume the Connections are called Transport Slice.

…which is fine except that you have moved the problem to the definition of “transport network”. The IETF will struggle, I think, with the idea that an IP network is a transport network unless, in the context of these documents, you give a very clear explanation of what these documents mean by that term.

 

But perhaps we can cut through all of this with some simple clarity. Text to describe the context and meaning of ‘transport’. Since the terminology document has so cheerfully plundered the enhanced VPN framework draft for some text, you might look there (in the Introduction) for suitable context-setting text.

 

That, of course, leads me to a separate question that I have, which is why the design team is reproducing and/or copying material from an existing WG document rather than working with that document to refine it and/or split it into multiple documents. A different question, but one the chairs might like to comment on.

 

Best,

Adrian

 

From: Teas-ns-dt <[e-mail] <mailto:[e-mail]> > On Behalf Of Rokui, Reza (Nokia - CA/Ottawa)
Sent: 30 April 2020 16:23
To: [e-mail] <mailto:[e-mail]> ; [e-mail] <mailto:[e-mail]> 
Cc: Rokui, Reza (Nokia - CA/Ottawa) <[e-mail] <mailto:[e-mail]> >
Subject: [Teas-ns-dt] FW: The word "transport"

 

All,

 

I thought I sent this to TEAS and TEAS-NS-DT team before. Forwarding the response to everyone.

 

Cheers,

Reza

 

 

 

From: Reza Rokui <[e-mail] <mailto:[e-mail]> >
Date: Saturday, April 25, 2020 at 11:37 PM
To: Dhruv Dhody <[e-mail] <mailto:[e-mail]> >, Jari Arkko <[e-mail] <mailto:[e-mail]> >, "[e-mail] <mailto:[e-mail]> " <[e-mail] <mailto:[e-mail]> >
Cc: Reza Rokui <[e-mail] <mailto:[e-mail]> >
Subject: Re: The word "transport"

 

Hi all,

 

Please see inline for some clarifications.

 

Reza

 

 

On 2020-04-24, 1:31 AM, "Dhruv Dhody" <[e-mail] <mailto:[e-mail]> > wrote:

 

    Hi Reza,

 

    I am putting the relevant text here -

 

    ==

 

    [14:28:42] <adrianfarrel> I'm getting very confused by everyone

    talking about "transport networks" and "transport slices". Is this

    something coming out of 3GPP?

 

[Reza] Please note that the “Transport Slice” from IETF point of view is not only limited to 5G network slicing.

5G network slicing is just one use-case of transport slicing as described in Transport Slice definition draft.

 

Regarding 5G use-case, the following figure is taken from 3GPP TS 28.530 where the “Transport slice” is described as “Connectivity” across transport network. 

This is the reason that what has been described in our draft for “Transport Slices” is a set of Connections between various endpoints with certain SLOs.

Note that 3GPP did not define the Transport slice explicitly.

 



 

 

 

    [14:29:06] <adrianfarrel> Seems to me that we (for example, ACTN) went

    through a lot to clarify that "transport network" was sub-IP

    [14:29:17] <adrianfarrel> ...maybe even sub-MPLS

[Reza] Transport networks (which are technology specific) are used to realized “Transport Slices”. Please see draft for more info.

 

    [14:31:39] <Joel Halpern> "Transport Network" is the industry

    terminology for the network that connects the radio (and related gear)

    to the packet core (and related gear).

    [14:32:03] <Joel Halpern> It is indeed a different usage.  I wish it

    were not overloading.  But they didn't ask me.

    [14:32:47] <Joel Halpern> Which part it refers to in a fixed access

    network is even less clear.

 

[Reza] The following figure is taken from 3GPP TS 28.530 where it shows various “Transport slices”. 

Note that the “Transport Slices” are not only used for RAN to Core. Depends on the deployment, we can have 

Transport slices in RAN for midhaul and fronthaul, in 5G Core or for application. All these transport slices are shown below.

 

] 

 

    [14:39:54] <adrianfarrel> I understand why 3GPP uses the term. I don't

    understand why we use the term.. We could talk about 'Foobar slices'

    and add one line that says "In the 3GPP, the term 'Transport Slice' is

    used for what we call a 'Foobar Slice'."

[Reza]  Note that 3GPP used the term “slice subnet” when describing it in context of 5G Core and RAN (i.e Core Slice Subnet and RAN Slice Subnet).

They did not define anything related to Transport. As indicated above, 3GPP refer it to “Connectivity”.

We did not want to use Subnet because in IP,  subnet has a very clear definition and using term “Transport Subnet” is completely misleading. So, we chose Transport Slice.

Also since the connectivity between various endpoints are across transport network (i.e. IP, Optics, PON, Microware etc.), it is logical to assume the Connections are called Transport Slice.

 

 

    [14:40:18] <adrianfarrel> We *already* have two definitions of

    "transport" in the IETF. We really don't need three

[Reza] Please provide links to these two definition.

 

    [14:59:00] <adrianfarrel> Well, I picked "foobar' in order to not jump

    immediately to a strawman solution. I think we are talking about

    providing slices as a service across the Internet. Same level of entry

    as a VPN: that is, the consumer provides a packet stream to the

    service entry point, and expects the packets to be delivered to the

    service exit point. Obviously, the consumer of the service may see the

    service as a transport (cf. pseudowires), but we would be 'alarmed' to

    hear a L3VPN described as a 'transport VPN.'

[Reza] VPN is one of the  solutions to realize the transport slices in IP networks.

However, The idea of Transport Slice is to allow a high-level system (or consumer or an Orchestrator) to ask for a various connections (i.e. Transport slices) across  IP, Optics, PON, Microwave or any combination of these networks. We shall not limit ourselves to IP VPN.

Note that the definition of the Transport slice is technology agnostic. For example in 5G you can realize a transpot slice in midhaul (please refer to picture above) which is a set of connections between 5G RAN nodes using PON or Optical connection. In this case there is no IP VPN. 

In summary, since the connectivity between various endpoints are across transport network (i.e. IP, Optics, PON, Microware etc.), it is logical to assume the Connections are called Transport Slice.

 

    Full logs - https://www.ietf.org/jabber/logs/teas/2020-04-23.html <https://protect2.fireeye.com/v1/url?k=d4f6bc46-8a560628-d4f6fcdd-86b1886cfa64-dd8e4ff74c5dda87&q=1&e=b15230b3-e6dc-4732-83e3-00354fdc75ae&u=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Fwww.ietf.org%2Fjabber%2Flogs%2Fteas%2F2020-04-23.html__%3B%21%21NEt6yMaO-gk%21WrvQt2QtERG2EFnE6CPHr8TzNjoFggUjNzUFGwWmlw2KiRSR42MiO2N4YtwhZJI%24> 

 

    ==

 

    More inline..

 

    On Thu, Apr 23, 2020 at 10:27 PM Rokui, Reza (Nokia - CA/Ottawa)

    <[e-mail] <mailto:[e-mail]> > wrote:

    >

    > Hi Dhruv,

    >

    > As far as I remember, we did not have any argument about term "transport". It was agreed that the term "Transport" is suited in context of network slicing.

    > The initial discussion was "Transport network slicing" which later we agreed to remove "network" since it causes confusion since IETF do not cover the e2e network slicing but rather the transport part.

    >

    > I was not clear about Adrian's comment regarding term "transport". IMHO, this term is well suited since it convers any underlying technology for L0 to L3.

    >

 

    In CCAMP/TEAS circles, the transport network term might not be used

    for L3 (and thus the confusion). And then there are the Transport

    Protocols! Definition draft should include some clarity on what do we

    mean when we say transport network at the least!

 

    It would be good to get this resolved sooner as we develop multiple

    documents that uses the same terminology!

 

    Thanks!

    Dhruv

 

    > Reza

    >

    >

    > On 2020-04-23, 12:35 PM, "Dhruv Dhody" <[e-mail] <mailto:[e-mail]> > wrote:

    >

    >     Hi,

    >

    >     Adrian gave some comments on jabber about using the word "transport".

    >

    >     During RFC8453 devolpment, this came up where TN in ACTN was transport

    >     networks and it was changed to TE networks to avoid using the

    >     overloaded term "transport".

    >

    >     We need to find a way to justify using this term and maybe describe

    >     this more in the definition draft - and just saying that 3GPP uses

    >     that term may not work I guess!

    >

    >     Was this discussed in the design team earlier, i joined the effort

    >     much later.

    >

    >     Thanks!

    >     Dhruv

    >

 

Juniper Business Use Only