[RFE] Define and set new network configurations

[QuantumPosix]

Currently QUADS only offers qinq 0 or qinq1

Having the capability to expand current offerings to include some different and needed network configurations:

New Additions
roce - Needed for 100G / 400G interfaces for testing of GPUdirect , StorageDirect
nvme - Connections to storage arrays for performance testing
iscsi - Same ^
lacp - example of what was asked of us to configure in the past for telco (and still active in current assignments) (also has this rfe - #631 )

[sadsfae]

New Additions
roce - Needed for 100G / 400G interfaces for testing of GPUdirect , StorageDirect
nvme - Connections to storage arrays for performance testing
iscsi - Same ^

I think we don't need to care about what the network is used for and thus try to build out extra complexity to categorize them.
Is this a setup that will be spun up and torn down regularly? If not perhaps we consider it a fixture and its not something that needs to be QUADS-managed.

For connecting hosts which probably do come and go, we can instead support extended designs to match the VLAN design(s) of a permanent fixture service.

Supporting more VLAN design modes can be tackled in one of the other issues you opened.

If you're concerned about mixing disparate hosts with varying network link speeds in a multi-tenant cloud then that's a jjob for --mod-switch-conf

[QuantumPosix]

This is specifically for the changes necessary for the switch configuration (juniper.py or other switch type) and other base quads changes needed for the other RFE's to work properly.
The new network types will be reusable for all hosts and can be added / removed easier with the other RFE's.

[sadsfae]

This is specifically for the changes necessary for the switch configuration (juniper.py or other switch type) and other base quads changes needed for the other RFE's to work properly. The new network types will be reusable for all hosts and can be added / removed easier with the other RFE's.

I got it. You have a couple of issues opened that sort of overlap with each other in some ways and I think can be brought into the same set of changes once it's discussed more, vetted and planned.

#684 mentions --define-network which can bridge a gap of it not really being easy to add more VLAN design modes. In fact not all labs need to have or support the same ones if they don't use them. We can leverage a supporting command-set like this to make it easier to define Q-in-Q additions per-lab.