Minutes IETF103: qirg
Quantum Internet Research Group
||Minutes IETF103: qirg
Quantum Internet Research Group (QIRG)
Chair: Rod van Meter (Rdv)
Thursday November 8th, 2018
• Quantum Internet, Axel Dahlberg (AD), 15 min.
Aaron Falk (AF): Is there a zeroeth order network where you are using binary
information to communicate between quantum computers and is that a useful thing?
AD: Possibly, wouldn't let you do anything more than a quantum computer could.
AF: Trying to understand if there is value in other points in the space.
AD: I don't know that there is anything that would allow you to do beyond just
having a quantum computer. A quantum computer will not be built before any of
these other things will happen.
Rdv: There is a proof that if you just take 2 quantum computers and connect
them only with a classical channel you get no net improvement - there's no big
gain. Q to Q communication is required to scale Q computation.
Igor Bryskin (IB): You are talking about entanglement but you haven't mentioned
decoherence - can you maintain entanglement and if so how?
AD: Will discuss briefly in next talk. There are things you can do if you can
only produce entanglement with a certain fidelity or a certain amount of noise.
You can also distil entanglement. Two bad entangled links can be combined to
something better. So you can improve entanglement even if you start with
something noisy. Storage can also suffer from decoherence - quantum memory is
inherently noisy. Part of the 5th stage - building gates in a fault tolerant
manner. It's a very important question but I don't have time now.
• Advertising Entanglement Capabilities in Quantum Networks
(draft-kaws-qirg-advent-00), Axel Dahlberg (AD), 15 min.
Dino Farinacci (DF): If you change an entangled qubit it's state will change at
both locations. Is that a true statement?
AD: State will change but it won't change the statistics of measuring other
qubit but will change correlations of measurement outcomes. This means that
there's no information being sent because from point of view of other qbit,
looked at locally, the probability of certain measurement outcomes doesn't
change, but the correlation changes.
Rdv: I'm a little fuzzy on details of the process. Draft has label qirg but
group hasn't taken on responsibility for it.
Aaron Falk: As a former IRTF chair, I can advise you that as the chair of the
RG you can run it in any way you see fit. Choose a process that's going to work
for keeping people engaged and making progress and try to achieve the goals
that you want.
Rdv: I'm perfectly happy to have it as the first item on our agenda. The other
co-chair is one of the co-authors on the draft so presumably she approves too.
Dino F: To move these properties we would need this quantum network. We'd need
to understand more fundamentals before we can know how to do things. If there's
no value to having a classical Internet between two quantum computers then why
are we designing to put quantum properties in a routing protocol that runs on a
AD: You have classical communication between nodes together with quantum
Dino F: Shouldn't we shoot for the goal of quantum communication only?
Rdv: That's one angle of what's being proposed. This is classical routing to
control the quantum network.
Alistair Woodman (AW): We're trapped in a corner case of quantum networking for
quantum computing - more interesting question is what is the value of quantum
networking to classical computing? Key distribution? Faster throughput? Use
cases could get people in this room interested, without the corner case.
AD: Proposal of different stages is also in an article in Science so look there
to understand more.
Rdv: QKD is the obvious example. It's an application for classical
communications that uses a minimal set of capabilities in the quantum world.
Part of the work item that's on the agenda is to try to figure out the whole
set of steps that go with it.
Igor Bryskin (IB): Proposal is similar to having a control plane that controls
an optical network. In this case we have a control plane that controls
communication between nodes based on properties of quantum links. Knowledge
will make it possible to create quantum links and to establish end-to-end
entanglement between two nodes. Is this an accurate description?
AD: Yes, I think so if I understood you. We will definitely need classiical
communication between nodes to build a quantum network, if only to synchronise
nodes. This proposal is simply to advertise capabilities of each link to make
IB: This is like how we control optical data plane. We do exchange routing
information of optical links to produce an optical trail with certain
characteristics along the trail. To produce entanglement between two directly
or not directly connected nodes I would need to know topology and set up
entanglement in such a way that it will work.
Dan Bogdanovic: I am confused with this draft. Much more interesting thing to
me is how to carry more information in a single photon over the network. Could
increase capacity - could end up being cheaper than how we do optical
networking today. Using classical networks to distribute information, whatever
that information is, we've been doing that.
Robert Broberg (RB): You might want to use QKD as an example with trusted nodes
and untrusted nodes - that might help to educate people about the simplest
thing you could do with a quantum network. We already have concepts of using
TLVs to find best path. Using this in a quantum space is pretty simple, but
people are having a hard time understanding why.
Rdv: I had that slide in the previous presentation about trusted repeaters -
did that not come through?
RB: No, I don't think so. People think of Beijing to Shanghai across multiple
hops, but that isn't it.
Melchior Aelmans: I'm one of the co-authors of this draft. Seems we need to be
clearer about our objectives in the draft. Our intention is to distribute the
data we need in order to establish entanglements. I agree that ideally we
wouldn't need extra links or a classical network but the fact is that at this
time we do need that.
• Draft charter, Rod Van Meter, 15 min.
Rdv reviewed the draft charter that had been shared with the mailing list
A Quantum Internet, if developed and deployed, will bring new
communication and remote computation capabilities, as well as
improving the accuracy of physical sensor systems (e.g., for
interferometry for long-baseline telescopes). One key area will be
cryptographic functions including quantum key distribution and quantum
Work toward a Quantum Internet is well underway in physics
laboratories and in theory groups. The next step is network
engineering. Some of the problems that need to be addressed include:
* routing: there are a number of proposals, including a couple in the
last six months or so, and which routing schemes are appropriate for
which circumstances needs to be assessed
* resource allocation: some of the routing proposals seem to be
including a notion of the dynamic traffic on the network, but this
distinction needs to be defined clearly
* connection establishment: what does a request look like (semantics
more than syntax) as it propagates across the network?
* interoperability: given than different networks are currently being
designed and built, how do we ensure a long-lived internetwork
* security: are quantum repeater networks inherently more or less
vulnerable in operations than classical networks?
* design of an API that will serve the role that sockets play in
There are also other problems:
* applications for a Quantum Internet: by far the most important on
the agenda for the community (not necessarily a QIRG work item) is
figuring out what we would *do* with a Quantum Internet, including
what data rates and fidelities are required
(otherwise, there is no market for a QI)
* multi-party states and multi-party transfers such as network coding:
rather than simple, independent point-to-point transfers, how can we
create and use more complex states?
Outputs and Milestones
Two concrete work items that QIRG may produce:
* An architectural framework delineating network node roles and
definitions, to build a common vocabulary and serve as the first
step toward a quantum network architecture.
* Wehner, Elkhouss and Hanson have created a roadmap of technical
capability milestones for quantum networks. Mapping these
milestones to concrete use cases will help to determine the order
and timing of classical protocols that will be needed.
Additionally, QIRG may serve as technical consultants to cryptography
groups within IETF and IRTF on the capabilities and timing for both
quantum computation and quantum networking, to provide guidance on
development and deployment of IETF protocols.
Finally, QIRG will serve as a coordination point with other standards
organizations that are working on standardization of quantum
QIRG will hold 2-3 meetings per year, typically one at IETF, one at a
related conference, and one online.
Aaron Falk: One of the questions I think that you have to ask when creating a
new research group is will its existence bring in the communities you will need
to have in order to collaborate. Clearly have gotten the interest of the IETF
community, 100 people in the room, but will you get the interest of people with
domain expertise that isn't typically at IETF. Have those people shown interest
- are you going to get them in the room?
Rdv: The physicists and experimentalists that are building these networks do
not have the kind of knowledge that is at IETF. It's not clear that they
recognise it. The EU quantum internet alliance effort that Axel described is
the largest effort of its kind so it is critical that we have their
involvement. Axel is a theorist and a software guy. Stephanie was at a European
ISP before moving into quantum. Other key part of Delft group is Ronald Hansen
- leading experimentalist - sits in same building with Stephanie and Axel -
collabroate on at least one key hardware project and path for this group to
influence that is through Axel and Stephanie. It's sort of a second class route
- would be better to have them show up in person and contribute to the mailing
list. So a key goal if we're going to meet in Prague, we need to have at least
one experimentalist in the room.
DF: That group of researchers need to teach us - we need to understand the
tradeoffs. Likewise we need to teach networking.
Rdv: I mentioned in my opening remarks the 3rd workshop for quantum networks
and repeaters. The first one was in 2015 where we started with a half-day
tutorial. We would be totally onboard with doing that again. We could do a
half-day tutorial in Prague.
** Show of hands who would turn out for that - 80% of the room **
Mikael Abrahamsson: I agree with the need for education. The analogy about
controlling an optical network seems close. We don't need to understand deep
physical detail, but do need to understand the properties and desired outcomes.
If you just want a routing protocol if you can elaborate on the requirements.
We can identify relevant routing protocols, need to understand enough about
properties and requirements, not necessarily deep physics. Analogies will be
Erik Nordmark: At the end of the proposed charter you talk about 'coordination
point with other standards organisations'. Inviting people to participate from
a research perspective is fine. You don't want to be receiving liaisons with
expectations of a response - not something research groups typically do.
Rdv: I want the weakest possible word that would include this. Maybe weakest
possible thing is to delete it. IEEE and ITU have some started some work on
this. Nothing that's as far forward as what Axel has been talking about. IEEE
group is working openflow control of QKD flows on trusted repeater networks -
an IEEE project working for about two years on that.
Alistair Woodman: Is it bandwidth, is it latency, people need a clear steer
with regard to what to expect at what time. People here turned up to understand
what's going, and maybe a tutorial is the way forward. We need greater clarity
on when this will inflect and on what applications. Not clear what you need
Rdv: Thanks for bringing that up. If you're thinking that quantum is a possible
way to get higher bandwidth out of your networks, you're thinking wrong -
that's not on the agenda. Data rates people talking about so far are as high as
maybe a Mb/s. For entangled networks, group in Delft are getting data rates of
30 entangled states per second. So the point of doing this is not bandwidth.
The point is that with entangled states you can do certain things in new ways
that circumvent that need for more bandwidth or fewer rounds of communication.
I divided the applications into three areas:
1. distributed computing
2. cryptographic protocols including QKD - point is to get new security
capabilties that you don't have in existing software systems. Similarly quantum
byzantine agreement - maybe could get better security properties than existing
cryptographic-based byzantine agreement protocols. open question if this will
deliver. 3. enhancing the precision of sensor networks including interferometry
between large radio or optical telescopes for example.
None of this is about bandwidth.
IB: I am specialist in optical control plane - I have limited understanding of
optical data plane. We run optimisations based on certain knowledge of optical
network without full understanding of what is happening in optical data plane -
this could be a similar paradigm to the relationship between IETF protocols and
the underlying physical properties of quantum networks. We will need to
understand the underlying processes but can also take advantage that we have so
Stewart Bryant: I'm a novice, but 2 observations - the time transfer solution
is one that is taxing the network industry - some of the new radios (e.g. 5G
and beyond) have critical requirements on exact time transfer that stress GPS
to its limit. And time is not about high bandwidth. Secondly, I observed a
paradox in what was being described - if improving security is a function of
quantum networking, how do we use a classical network to interconnect nodes if
they are vulnerable to classical security problems?
Rdv: That is the million dollar question. Part of the answer is that in fact
some of these security protocols that depend on end-to-end entanglement allow
the two end nodes to confirm with each other that they believe that there is
no-one in the communication channel but if their classical communication can be
disrupted in real time including the breaking of whatever encryption they're
using for that then yes you've got a chicken and egg problem. It's an open
• AOB/open mic, 10 min.
Rdv: Please join the mailing list, there are reference materials including for
beginners in the mailing list archives.