Agenda
1. Administrivia (5 min):
Resolving any online issues
RG update
[WK]: two RFCs published since last meeting,
Please review and comment on draft-lopez-qirg-multiplane-arch.
[Diego Lopez]: draft plays with ideas to analyze proposals and work toward an architecture framework. Only thing new from prior version is material on applying network virtualization ideas. Planning a -03 version prior to Bangkok meeting
[Tim Chown]: Likely at some point research networks will be dploying this technology. Where can we look for deployment guidelknes, how to operatie, how research netwoks can use leased fiber, etc.?
[DL]: one of our roles here is to learn this at small scale first. in Madrid have 4 nodes - pretty far away form a real network.
[MC]: Once you create entanglement among remote nodes, you have classical communication that exceeds the boundary of your locall network. So you need a classical routing protocol in place, and inter-network forwarding in the classical domain.
[TC]: where in this is the operations and management side of it. Anybody looking into this? How do you manage it
[MC]: you ask what the classical signaling you ise to do this. In the lab we need to start impleneting these protocols based on TCP or UDP. Currently labs are re-inventing the wheel, bcasue the lab people don't have the networkingbackground & expertise. Example: time synchronization - most labs not using current protoclls for this. Have to be sure there is collaboration.
[DL]: this is what we are trying to achieve by having a playground fo both sets of people to work together. In the end you have to identify entangled state - in the netowkring sense this seems to be the equivalent of addressing? What does "quantum addressing" look like
[MC]: hope to have more to say on quantum state addressing for next meeting, but first you have to undertand the use cases for whether you need hierarchical addressing or not.
[Roland Bless]: is multipartite entanglement more fragile than simple two-state entangement?
[MC]: you can do this within a single node; current quantum chips are already doing this in a central node. Then you do EPR and teleport the entanglement to the other nodes. Hard part is converting the in-chip entanglement to photons that can go between nodes.
[Tim Chown] this is fantastic operational commentary. From an oprational point of view this is exactly what you need. Q: did you have to bing in extra expertise, or did the existing engineers learn without external help?
[PR]: need to have good contacts to help
[Wei Pan]: if the QKD processors can be a module that can be plugged into regular network devices
[PR]: you co-propagate the quantum data together with the classical channel, but there are distance limitations.
[IM]: maybe IEEE, maybe a new quantum-specific organization.
[Björn Haase]: I know that some early designs assigned maybe one qubit for addition and sixty for communication, what is your current thinking on this?
[IM]: some qubits will need to be reserved for communication compared to those needed for computation. Since the communication qbits are consumed, you have to regenerate them and there are open issues around time/space tradeoffs
[Craig Rodine]: are these single photons or streams?
[IM] on 1550nm we send multiple ones since the efficiency is 50% or less. Can't use EDF amplifiers
[rdv]: what's the rate they are generated/sent?
[IM]: don't have good numbers yet, there are losses at each hop. Now doing trials at 10K/s
[WK]: are you running any applcations yet?
[IM]: We saw good discussion of applications at IEEE Quantum Week in Montreal (in September). We see a happy marriage emerging. First "application" will just be a CNOT gate.
[Dave Oran]: RT scheduler -- what kind of time frame? Nanosecond? Millisecond? How big are the the time slots?
[WW]: Time slots are 10usec.
[DO]: You need clock sync about an order of magnitude better than that.
[WW]: Others can be faster.
[DO]: Do you have to build an accurate level of sync on the link level? If your time slot isn't accurate, you might lose sync? Do you have to do this when the photons arrive?
[WW]: No. We do 20usec ?post-processed?, but others are picosecond.
[Diego Lopez]: This NBI, northbound interface?
[WW]: All layer one, layer two.
[DL]: What NBI are you going to use?
[IM]: we're not there yet on an API for northbound to applications
[DL] have to design, name and identify the abtract objects that are to represent the varoious parts of the system, things like adddresses, channels, etc.
[WW]: maybe the socket abstraction is right - just name the sockets and "inside" them are the matter and copmmuniction Qbits that arrive, and then may decohere.
[WK] was/is a fan of using YANG models and that's what people are trying to migrarte to.