I think the concept of an InterfaceConnection could be extended to have a Characteristic (Virtual or Physical) along with a "generic user-defined type" where the user can define any number of types they want along with custom fields for each. This way NetBox isn't trying to model everything, but just provides a generic framework for connecting interfaces to a "connection" and allowing generic types to be defined on that connection to account for everyone's specific case.
You'd have to have another table/model for the actual attachment definitions such as InterfaceConnectionAttachment that would actually specify which connection is attached to which interface along with a the "type" of attachment it is (probably Root or Leaf).
So you could allow any given connection to be tied to 2 *or more* interfaces and loosen the requirement that it must be one-to-one. This would allow users to represent say a wireless Access Point that has 10 subscribers linked to it. This example would be one "connection" w/ the AP wireless interface attached as the "root" and each subscriber unit interface attached as "leaves". This is an example of a "physical" connection that feels virtual. It's physically linked via RF, but can't be modeled in NetBox. One could also attach one UNI to another remote UNI or NNI and mark it as a "virtual" type and give it their own type such as "EVC" to represent an ethernet-virtual-circuit along with S-VLAN and C-VLAN custom fields.
Or like Simon wants -- to represent one interface with hundreds of vlans on it. That one interface could have 100 "VLAN"-typed "connections" on it .
To summarize, I think virtual connections could be modeled fairly easily, but it's a drastic change from the one-to-one physical-only model there is now.
- Make the concept of a "connection" and an "interfaceconnectionattachment" so that connections can have many interfaces attached (multi-point) or an interface can have multiple connections on it.
- When attaching a connection to an interface, you'd have to specify if that connection is a Root or a Leaf. Examples...
- If it's a p2p, both would be Root.
- If it's a wireless AP, the AP interface would be Root and clients would be Leaf.
- If it's an ELAN, everyone would be Root.
- If it's an ETREE, only the root ports would be root and other UNI ports Leaf.
This would be a key requirement otherwise NetBox wouldn't know how to render a bunch of connections/interfaces. It would have no frame of reference on where the "root" is and if you had a concept like the Wireless AP w/ clients linked, you would want the system to know which is the AP and which is the sub so you can render the relationship that way.
- Add a characteristic to the "connection" so that one can specify it as "Virtual" and attach remote interfaces together virtually, or even in a multi-point fashion as well to represent concepts like ELAN (VPLS).
- Allow users to define custom "Connection Types" each with their own fields. This would allow users to add any number of "types" they want like "Fiber", "Copper", "EVC", "Wireless", etc each with their own set of relevant fields for the connection.
On Thursday, January 18, 2018 at 10:14:13 AM UTC-7, Jeremy Stretch wrote:
NetBox doesn't yet support virtual circuits, and frankly it probably won't for quite some time given the difficulty in modeling them. (A virtual circuit can take many different forms.)
Jeremy
Is there anyway to change this? As it would be a very handy feature, due to the carriers now using 10G aggregated handoffs
On Friday, 12 January 2018 14:31:32 UTC, Brian Candler wrote:Ah I see.
From the docs: "Each circuit termination is tied to a site, and optionally to a specific device and interface within that site." But I've tested it, and multiple Circuits cannot be linked to the *same* device and interface.
So the best Netbox can do is represent the physical bearer, but not the individual circuits delivered over that bearer :-(
--
You received this message because you are subscribed to the Google Groups "NetBox" group.