Definitions of Topology, Domain and Network
Hello, As agreed at the last OGF, Inder and I have worked on the definitions of Network, Topology and Domain. - Topology: A connected graph of Network Elements The intended usage of this is to describe the thing that a network provider advertises to others as his network topology that is available for use. - Domain: An unordered collection of Network Elements managed under the same shared policy umbrella. This describes who or what is responsible for the Network Elements. - Network: no definition We had a discussion about the meaning of this term and in the end we had to conclude that is not possible to provide a workable definition for the term 'Network'. The term Network has become so widely used for so many diverse meanings that it is impossible to create a strict definition that everyone can agree on, while still expressing something useful. Inder & Jeroen.
On Dec 14, 2009, at 7:47 AM, Jeroen van der Ham wrote:
Hello,
As agreed at the last OGF, Inder and I have worked on the definitions of Network, Topology and Domain.
- Topology: A connected graph of Network Elements The intended usage of this is to describe the thing that a network provider advertises to others as his network topology that is available for use.
So, in a case where a network has two disjoint subsets (e.g. Northrop Grumman's two campuses), they'll advertise two separate topologies? Cheers, Aaron Winter 2010 ESCC/Internet2 Joint Techs Hosted by the University of Utah - Salt Lake City, UT January 31 - February 4, 2010 http://events.internet2.edu/2010/jt-slc/
On 14/12/2009 15:05, Aaron Brown wrote:
On Dec 14, 2009, at 7:47 AM, Jeroen van der Ham wrote:
Hello,
As agreed at the last OGF, Inder and I have worked on the definitions of Network, Topology and Domain.
- Topology: A connected graph of Network Elements The intended usage of this is to describe the thing that a network provider advertises to others as his network topology that is available for use.
So, in a case where a network has two disjoint subsets (e.g. Northrop Grumman's two campuses), they'll advertise two separate topologies?
Exactly. Jeroen.
On Dec 14, 2009, at 6:19 AM, Jeroen van der Ham wrote:
On 14/12/2009 15:05, Aaron Brown wrote:
On Dec 14, 2009, at 7:47 AM, Jeroen van der Ham wrote:
Hello,
As agreed at the last OGF, Inder and I have worked on the definitions of Network, Topology and Domain.
- Topology: A connected graph of Network Elements The intended usage of this is to describe the thing that a network provider advertises to others as his network topology that is available for use.
So, in a case where a network has two disjoint subsets (e.g. Northrop Grumman's two campuses), they'll advertise two separate topologies?
Exactly.
Jeroen.
To be the devil's advocate, this leads to a situation where, for example, a single GOLE that provides different services (i.e. lightpath and vlan and SDH with no translation/encapsulation/ multiplexing capabilities), will need to provide a separate "topology" per service, since the optical switch is not "connected" to the ethernet switch. Does that make sense? It looks unnecessarily complex to me. If we had the concept of a "connected subgraph" of a domain or topology, that might help with things.. a network provider would advertise a single topology object that would contain one or more of these.
On 14/12/2009 18:31, Evangelos Chaniotakis wrote:
To be the devil's advocate, this leads to a situation where, for example, a single GOLE that provides different services (i.e. lightpath and vlan and SDH with no translation/encapsulation/ multiplexing capabilities), will need to provide a separate "topology" per service, since the optical switch is not "connected" to the ethernet switch. Does that make sense? It looks unnecessarily complex to me.
You mean a GOLE that has an optical switch that is in no way connected to the ethernet switch, i.e. there is no cable running between them? Then I'd say that they are actually two different GOLEs.
If we had the concept of a "connected subgraph" of a domain or topology, that might help with things.. a network provider would advertise a single topology object that would contain one or more of these.
We have to break things down into manageable chunks somehow. This is one that seemed most natural. I'm sure there are also examples of a single topology that is provided by multiple providers. Jeroen.
I am wondering if it is possible to combine two definitions to come up with what you would like to see "If we had the concept of a "connected subgraph" of a domain or
topology, that might help with things.. "
for example: Domain Topology = connected subgraph of network elements within a domain. Physical Topology = connected subgraph defining the physical connections between the network elements .... Inder On Dec 15, 2009, at 6:32 AM, Jeroen van der Ham wrote:
On 14/12/2009 18:31, Evangelos Chaniotakis wrote:
To be the devil's advocate, this leads to a situation where, for example, a single GOLE that provides different services (i.e. lightpath and vlan and SDH with no translation/encapsulation/ multiplexing capabilities), will need to provide a separate "topology" per service, since the optical switch is not "connected" to the ethernet switch. Does that make sense? It looks unnecessarily complex to me.
You mean a GOLE that has an optical switch that is in no way connected to the ethernet switch, i.e. there is no cable running between them?
Then I'd say that they are actually two different GOLEs.
If we had the concept of a "connected subgraph" of a domain or topology, that might help with things.. a network provider would advertise a single topology object that would contain one or more of these.
We have to break things down into manageable chunks somehow. This is one that seemed most natural. I'm sure there are also examples of a single topology that is provided by multiple providers.
Jeroen. _______________________________________________ nml-wg mailing list nml-wg@ogf.org http://www.ogf.org/mailman/listinfo/nml-wg
On Dec 15, 2009, at 9:32 AM, Jeroen van der Ham wrote:
On 14/12/2009 18:31, Evangelos Chaniotakis wrote:
To be the devil's advocate, this leads to a situation where, for example, a single GOLE that provides different services (i.e. lightpath and vlan and SDH with no translation/encapsulation/ multiplexing capabilities), will need to provide a separate "topology" per service, since the optical switch is not "connected" to the ethernet switch. Does that make sense? It looks unnecessarily complex to me.
You mean a GOLE that has an optical switch that is in no way connected to the ethernet switch, i.e. there is no cable running between them?
Then I'd say that they are actually two different GOLEs.
If we had the concept of a "connected subgraph" of a domain or topology, that might help with things.. a network provider would advertise a single topology object that would contain one or more of these.
We have to break things down into manageable chunks somehow. This is one that seemed most natural. I'm sure there are also examples of a single topology that is provided by multiple providers.
A question still open is the definition of "connected"? Is it a literal connected graph, or does it mean connected such that folks could actually somehow make circuits to get from any point in the graph to any other point (ignoring how they know that reservations and the like can happen)? For example, say someone has a switch with sonet ports and ethernet ports and that switch connects to two other nodes, one via ethernet and one via sonet. Is the implication that the node connected via ethernet can connected to the node connected by SONET? If not, is that a connected graph for these purposes, or are there two separate topologies (the SONET one and the Ethernet one)? Relatedly, if a topology is disjoint due to solely to switching capabilities instead of cabling, is that two separate topologies or a single topology? Cheers, Aaron Winter 2010 ESCC/Internet2 Joint Techs Hosted by the University of Utah - Salt Lake City, UT January 31 - February 4, 2010 http://events.internet2.edu/2010/jt-slc/
On Dec 16, 2009, at 9:12 AM, Aaron Brown wrote:
A question still open is the definition of "connected"? Is it a literal connected graph, or does it mean connected such that folks could actually somehow make circuits to get from any point in the graph to any other point (ignoring how they know that reservations and the like can happen)?
This is a good question. In my mind being able to make circuits (or connections) implies that there is an encapsulation and "concatenation" capability at all points on the graph.
For example, say someone has a switch with sonet ports and ethernet ports and that switch connects to two other nodes, one via ethernet and one via sonet. Is the implication that the node connected via ethernet can connected to the node connected by SONET? If not, is that a connected graph for these purposes, or are there two separate topologies (the SONET one and the Ethernet one)?
I wonder if there needs to be a concept of "topology at a layer", where layer is a particular technology layer.
Relatedly, if a topology is disjoint due to solely to switching capabilities instead of cabling, is that two separate topologies or a single topology? I am not sure what you mean by this -- how would switching capabilities make something disjoint?
John
On Dec 16, 2009, at 10:54 AM, John Vollbrecht wrote:
Relatedly, if a topology is disjoint due to solely to switching capabilities instead of cabling, is that two separate topologies or a single topology? I am not sure what you mean by this -- how would switching capabilities make something disjoint?
Due to hardware limitations or the like, only certain ports can be cross-connected together. If you end up with a disjoint sets of ports that can be cross-connected, you could end up with an overall disjoint network (especially if it's a GOLE and you only have one or two switches). Cheers, Aaron Winter 2010 ESCC/Internet2 Joint Techs Hosted by the University of Utah - Salt Lake City, UT January 31 - February 4, 2010 http://events.internet2.edu/2010/jt-slc/
On 16/12/2009 15:12, Aaron Brown wrote:
A question still open is the definition of "connected"? Is it a literal connected graph, or does it mean connected such that folks could actually somehow make circuits to get from any point in the graph to any other point (ignoring how they know that reservations and the like can happen)?
For example, say someone has a switch with sonet ports and ethernet ports and that switch connects to two other nodes, one via ethernet and one via sonet. Is the implication that the node connected via ethernet can connected to the node connected by SONET? If not, is that a connected graph for these purposes, or are there two separate topologies (the SONET one and the Ethernet one)? Relatedly, if a topology is disjoint due to solely to switching capabilities instead of cabling, is that two separate topologies or a single topology?
I think that "connected" is indeed a term that we should define properly. I would say that for "connectedness" it has to be possible for a circuit to be made between ports ot make it "connected". It does not have to be a terminated connection, so it should be possible to come in on Ethernet on one side and SONET on the other. Jeroen.
Aaron Brown wrote:
A question still open is the definition of "connected"? Is it a literal connected graph, or does it mean connected such that folks could actually somehow make circuits to get from any point in the graph to any other point (ignoring how they know that reservations and the like can happen)?
For example, say someone has a switch with sonet ports and ethernet ports and that switch connects to two other nodes, one via ethernet and one via sonet. Is the implication that the node connected via ethernet can connected to the node connected by SONET? If not, is that a connected graph for these purposes, or are there two separate topologies (the SONET one and the Ethernet one)?
Very good point. Thanks. I'm also puzzled about the topology definition:
- Topology: A connected graph of Network Elements The intended usage of this is to describe the thing that a network provider advertises to others as his network topology that is available for use.
I do not know what a "graph of Network Elements" is. My question is the same as Aaron's: is a network with two (physically connected) devices (a Ethernet and a SONET device) one or two graphs? The answer presumably depends on how the (connected) Network Elements are mapped onto a graph: as a graph representation of physical devices; a graph representation of abstracted or logical devices; (multiple) graphs for each logical layer? Instead of going in loops to define the word "graph", I would suggest to simply remove the word graph from the definition:
- Topology: A connected set of Network Elements
Regards, Freek
I like these definitions, they seem to fit the model of transport and service plane. On Dec 14, 2009, at 7:47 AM, Jeroen van der Ham wrote:
Hello,
As agreed at the last OGF, Inder and I have worked on the definitions of Network, Topology and Domain.
- Topology: A connected graph of Network Elements The intended usage of this is to describe the thing that a network provider advertises to others as his network topology that is available for use. I assume that a topology could include a set of interconnected groups, where that topology is what is available to others from the set of groups. Can a topology be a set of ports that are potentially interconnected? This sounds like a description of transport plane.
- Domain: An unordered collection of Network Elements managed under the same shared policy umbrella. This describes who or what is responsible for the Network Elements.
This sounds like what NSI is calling a NS actor. In the case where a NS actor calls other NS actors we might need a pseudo domain concept for the group of NSAs providing resources.
- Network: no definition We had a discussion about the meaning of this term and in the end we had to conclude that is not possible to provide a workable definition for the term 'Network'. The term Network has become so widely used for so many diverse meanings that it is impossible to create a strict definition that everyone can agree on, while still expressing something useful.
I agree with this.
Inder & Jeroen. _______________________________________________ nml-wg mailing list nml-wg@ogf.org http://www.ogf.org/mailman/listinfo/nml-wg
Jeroen van der Ham wrote:
As agreed at the last OGF, Inder and I have worked on the definitions of Network, Topology and Domain.
Inder and Jeroen, Thank you for your contribution on the terms "Topology" and "Domain". Could I ask you to take step two and refine the definitions with the feedback from the group members? As I understand, the open questions are: - Topology: How is "connected" defined? - Topology: you refer to "graph". What are the nodes and edges in this graph? (or did you mean to say "collection" instead?) - Domain: It is defined a collection of *network* elements, so a collection of both network and computing elements is *not* a domain. Is that intended? - Is it correct that the distinction between topology and domain is the same distinction as transport and service plane? Regards, Freek
Hello, We took your comments, went back to our definitions and updated some of them. The new definitions are below: Topology: A set of Network Elements and the links connecting them NetworkDomain: An unordered collection of Network Elements managed under the same shared policy umbrella. To make the distinction clearer that we are describing network elements only, we're calling this a NetworkDomain instead of a regular Domain. Jeroen. On 14/12/2009 05:47, Jeroen van der Ham wrote:
Hello,
As agreed at the last OGF, Inder and I have worked on the definitions of Network, Topology and Domain.
- Topology: A connected graph of Network Elements The intended usage of this is to describe the thing that a network provider advertises to others as his network topology that is available for use.
- Domain: An unordered collection of Network Elements managed under the same shared policy umbrella. This describes who or what is responsible for the Network Elements.
- Network: no definition We had a discussion about the meaning of this term and in the end we had to conclude that is not possible to provide a workable definition for the term 'Network'. The term Network has become so widely used for so many diverse meanings that it is impossible to create a strict definition that everyone can agree on, while still expressing something useful.
Inder& Jeroen. _______________________________________________ nml-wg mailing list nml-wg@ogf.org http://www.ogf.org/mailman/listinfo/nml-wg
participants (6)
-
Aaron Brown -
Evangelos Chaniotakis -
Freek Dijkstra -
Inder Monga -
Jeroen van der Ham -
John Vollbrecht