Do you have the latest schema? I would like to understand how this all fits together. I have some questions about how multiple circuits can be carried over a Link. And how multiple paths can be carried over a sequence of links. Perhaps I can ask questions better after looking at the schema. John On Jul 1, 2008, at 8:24 AM, Jeroen van der Ham wrote:

John Vollbrecht wrote:

...

Of particular note are some terms we use that don't seem to be in the NML schema (perhaps I am wrong?)

path - a connection between a source and destination. A path is a sequence of hops. source - starting point of path - as defined by direction of signalling destination - end point of path- as defined by direction of signalling hop - a network element - [domain or node or port or link]

These are all in the NML schema as proposed so far. We have Path and NetworkElement objects. And I think the Source and Destination could be encoded as SourceDomain and DestinationDomain. The *Domains part has not been defined completely yet. But I think we can take Victors' presentation as a startingpoint for that?

...

path segment - subset of a path consisting of two or more hops circuit - a connection between tow endpoints that can be used to transmit data between them

In the discussion in Berlin it was also suggested that we use the term "Path" also for path segments. Whether the path is complete can then be checked by the using Source- and DestinationDomain.

As for the circuit, I think that that is something for a later schema, as it would require descriptions of layers. (otherwise you cannot determine whether you can transmit data over it).

Jeroen. _______________________________________________ nml-wg mailing list nml-wg@ogf.org http://www.ogf.org/mailman/listinfo/nml-wg

John Vollbrecht, Senior Network Engineer Internet2 office +1-734-352-4960 | mobile +1-734-395-7890

On Jul 1, 2008, at 9:09 AM, Freek Dijkstra wrote:

John Vollbrecht wrote:

...

Attached is a DCN glossary which is part of DCN architecture and DCN reservation documents.

Thanks John!

...

path - a connection between a source and destination. A path is a sequence of hops.

I would argue that a path is a sequence of edges/links rather than hops. That way I can distinguish between two links between the same hops, and can more easily between different types of relations between hops (link, cross connect, adaptation, de-adaptation)

I think of a path a a sequence of hops because a requested path may contain some but not all intermediate hops, and hops may be domains rather than links. A strict path is a sequence of links. There may be multiple paths between hops but paths may be "potential" or requested as well as confirmed and strict. What I think is needed is a concept of resources like bw that are included in a connection or circuit. I have (I think wrongly) been using path to describe a circuit, which is path with specific resources in each segment of the path. Probably we need a name for this concept of circuit over time. Perhaps just circuit is good enough -- what do you think. Then what DCN reserves is a circuit and its circuit segments. Paths and path segments are similar without resources assigned to them. does this make sense? In that case, adaptations are assigned to circuits rather than links. Which is not to say that circuits cannot be links in a different infrastructure of higher order links and nodes. John

Regards, Freek

John Vollbrecht, Senior Network Engineer Internet2 office +1-734-352-4960 | mobile +1-734-395-7890

I think Evangelos is right -- we are talking about slightly different things. I think there are several differences that might be considered 1) for me the difference between a "path" and a "strict path" I think that what Freek describes is a strict path in my terminology. It contains edgepoints (i.e. links), and contains all the (relevant) links A path is just a way to get between two edge points. It may be described in a number of ways a) just the two endpoints b) endpoints + some intermediate hops - where hops can be links or other topological element (i.e. domain, node, port) note that for me a path does not have to specifically define every link in the topology between end points - it is "potential" 2) Freek seems to assume that adaptations occur only at links. In my model the hops in a path may be domains, for example. The implication is that domains must be able to do adaptations. 3) My potential paths may assume adaptations are possible without knowing for sure that they are. I think the potential path probably needs a name different than the actual path 4) I have taken to using circuit as a path with specific resources assigned to it. It is possible, for example, to have multiple circuits using the same path. Circuits are what DCN creates over paths. Adaptations take place on circuits in my view, not paths. It seems to me paths have adaptation points in them but don't do the adaptation (perhaps this is not right - perhaps adaptations are part of paths and circuits -- ?) John On Jul 1, 2008, at 5:24 PM, Evangelos Chaniotakis wrote:

Hi,

My opinion is that we're talking about slightly different things.

Path in the sense Freek defines it seems to represent a "finalized" object while the way John describes it is closer to an "unfinished" object that we would find in a service request.

I think both approaches have value, and are in fact very close but maybe we would like to to differentiate between them?

On Jul 1, 2008, at 8:44 AM, John Vollbrecht wrote:

...

On Jul 1, 2008, at 9:09 AM, Freek Dijkstra wrote:

...

John Vollbrecht wrote:

...

Attached is a DCN glossary which is part of DCN architecture and DCN reservation documents.

Thanks John!

...

path - a connection between a source and destination. A path is a sequence of hops.

I would argue that a path is a sequence of edges/links rather than hops. That way I can distinguish between two links between the same hops, and can more easily between different types of relations between hops (link, cross connect, adaptation, de-adaptation)

I think of a path a a sequence of hops because a requested path may contain some but not all intermediate hops, and hops may be domains rather than links. A strict path is a sequence of links.

There may be multiple paths between hops but paths may be "potential" or requested as well as confirmed and strict.

John Vollbrecht, Senior Network Engineer Internet2 office +1-734-352-4960 | mobile +1-734-395-7890

Hello all of you, Evangelos Chaniotakis wrote:

I think both approaches have value, and are in fact very close but maybe we would like to to differentiate between them?

Related to this, is a point I try to make more explicit. In some case entities should have data for the actually configurated values and in some cases one needs to provide the alternative values a parameter can have (which could be provisioned). Perhaps one needs to distinguish here in different class or attrbiute: configuredValue possibleValue Remember in some cases the existing network topology is not very handy to determine a possible other network configuration. E.g. in the GLIF environment most links were made by adding new physical entities (interfaces, hops, etc.)... So this is the difference between provisioning and configuration management. In most cases (but not always) provisioning is a human activity and I think we will need to start realizing that such alternatives are possible. An example; multiple Ethernet tagging method can sometimes be configured through an interface (like for most switches: 802.3, 802.1Q, 802.1ad). So these options exist in that 'node'/'hop'. And the choice is sometimes depending on the type of service one needs to provide (e.g. one needs to do 802.1ad on a trunk if one wants to provide 802.1Q to clients). Another example; a network might advertise already its willingness that it can do more functions. But these will need human interventation (like an NREN is willing to change the interface to add the 802.1ah feature). Perhaps people see the above as futuristic (real SF), but I think that the NML should be able to cater for that idea which is part of the SF (so in some way it should allow multi-value fields). Hope I did not complicate things too much;-) All the best, Victor -- Victor Reijs, Network Development Manager HEAnet Limited, Ireland's Education and Research Network 1st Floor, 5 George's Dock, IFSC, Dublin 1 Registered in Ireland, no 275301 tel: +353-1-660 9040 fax: +353-1-660 3666 web: http://www.heanet.ie/

Ralph - Thanks for the discussion - I think this raises a question about how one thinks about paths and links. I think one difference between IP and dynamic circuits is that dynamic circuits require setting up ete circuits and reserving resources for the circuits. These reservations are done resource providers. We tend to think of a "domain path segment"' as being between provider end points. Other segments or paths could end, as you point out, at the incoming link on the remote equipment. Since either way the end of the path is on the same link, the question is why it makes a difference. If the end is on the provider side, then the provider is able to set up all parts of the circuit and monitor it. What is requested of a provider is something he can provide. I note that in our user interface we allow request for circuits to be made between devices (i.e. the user side of the link) and the user interface converts it to provider endpoint links before forwarding it to the provider. This allows requesting connection from user interface (as in IP) and then requesting from the provider what it is able to provide. For monitoring we assume that both sides of an interdomain link segment can be monitored by their respective domain. The link segment between user and provider (or provider and provider) has different names at each end of the link, and the topology must describe this. The reason for different names at each end is that the domain must be part of the name in each case. Thus interdomain links have two names for what (if all works correctly) is the same link. Monitoring at each end assures that the segment is not broken. I think perhaps there needs to be some qualification of what sort of path is required - i.e. a user - user path or an network provider path. What do you (and others) think? I think that once a provider path is complete, users at each end must also verify that the path is operational, at that point the path includes users. John On Jul 1, 2008, at 9:55 AM, Ralph Niederberger wrote:

Dear John and all,

only some information which could influence some definitions.

I had some discussion with GARR (Italian NREN) concerning Endpoint and Demarcation point of an E2E path. Background is the following:

GEANT2 provides wavelength throught their infrastructure which can be prolonged on both sides by wavelengths of NRENs. DEISA has rent such a kind of 10 Gbit/s End to End Service (e2e).

E.g. DEISA site FZJ Juelich, Germany <-> fiber cable < -> DWDM equipment at PoP of local NREN DFN in Juelich, Germany <-> fibre cable with DWDM <-> DWDM equipment of local NREN DFN at GEANT2 PoP in Frankfurt, Germany <-> fibre cable <-> DWDM equipment of GEANT2 at Frankfurt, Germany <-> fibre cable with DWDM <-> DWDM equipment of GEANT2 at London, UK <-> fibre cable <-> DWDM equipment of remote NREN JANET <-> fibre cable with DWDM <-> DWDM equipment at PoP of remote NREN JANET at Daresbury, UK <-> fibre cable <-> DEISA remote site EPCC Daresbury, UK

For monitoring purposes the "GEANT2 End2End Coordination Unit" (E2ECU) has setup a domain related monitoring tool. Each domain provides its information of the path (e.g. Path is administratively or operationaly up or not). Measurement points between domains are demarcation points. Both ends of the link between two domains provide partial information. So concerning to this definition

GARR argues that the end of the provider links are the end point of the path, i.e. the outgoing interfaces of the DWDM equipment.

IMHO the ends of the path are the incoming interfaces of the two sides which have requested the e2e path.

So your definition would give room for some interpretation: "path - a connection between a source and destination. A path is a sequence of hops." We all know, that we assume as path (at least within IP) the connection from source (outgoing interface) to destination (incoming interface). But others (e.g. Service providers) could argue in a different way. best regards

Ralph John Vollbrecht schrieb:

...

Attached is a DCN glossary which is part of DCN architecture and DCN reservation documents.

Of particular note are some terms we use that don't seem to be in the NML schema (perhaps I am wrong?)

path - a connection between a source and destination. A path is a sequence of hops. source - starting point of path - as defined by direction of signalling destination - end point of path- as defined by direction of signalling hop - a network element - [domain or node or port or link]

path segment - subset of a path consisting of two or more hops circuit - a connection between tow endpoints that can be used to transmit data between them

I note that in the glossary some information is assumed, in at least some of the definitions. In particular, a path (perhaps a DCN path) must include time/duration and resource. That is, a Link may carry multiple paths at a given time, and may carry different paths sequentially.

I am wondering if this or something like it can be incorporated (or is incorporated) in NML.

John

John Vollbrecht, Senior Network Engineer Internet2 office +1-734-352-4960 | mobile +1-734-395-7890

--------------------------------------------------------------------- ---

_______________________________________________ nml-wg mailing list nml-wg@ogf.org http://www.ogf.org/mailman/listinfo/nml-wg

--

*************************************************** Ralph Niederberger Juelich Supercomputing Centre Institute for Advanced Simulation

Phone: +49 2461 61-4772 Fax: +49 2461 61-6656 E-Mail: r.niederberger@fz-juelich.de WWW: http://www.fz-juelich.de/jsc/

JSC is the coordinator of the John von Neumann Institute for Computing and member of the Gauss Centre for Supercomputing ***************************************************

Forschungszentrum Jülich GmbH 52425 Jülich

Sitz der Gesellschaft: Jülich Eingetragen im Handelsregister des Amtsgerichts Düren Nr. HR B 3498 Vorsitzende des Aufsichtsrats: MinDirig'in Bärbel Brumme-Bothe Geschäftsführung: Prof. Dr. Achim Bachem (Vorsitzender), Dr. Ulrich Krafft (stellv. Vorsitzender), Prof. Dr. Harald Bolt, Dr. Sebastian M. Schmidt ***************************************************

John Vollbrecht, Senior Network Engineer Internet2 office +1-734-352-4960 | mobile +1-734-395-7890