OFC2024 Workshop
and Technical program presentations & involvement by
UvA group members.
Birds of a Feather: Designing
and Operating the Next Generation Optical Photonic
Networks
OFC2003 - Monday, March 6,
2023. 4:30 - 6:30 PM
SDCC Room 14B
- Lots of people, counted > 60
- Slides: all presentations in one
pdf file.
- Welcome, Introduction by Cees de Laat
- Slide 2. Idea of OFCnet was to bring demonstrations,
R&E and experimental networks, etc, to the
showfloor. Refer to OFC 2023 - OFCnet Architecture
Diagram and OFCnet Demonstrations for details on the
network (“OFCnet”) which was built to support various
showfloor demonstrations. Three 25km pairs of fiber out
to a commercial facility near a small airport north of
San Diego (aka “Aero Drive”).
- Slide 3. Goals for this BoF: Bridge between academia
& researchers with capabilities deployed on the
showfloor, and figure out what kinds of incentives will
work to attract interest. Is there a need for a workshop
series? Publications, what researchers or vendors need -
revisit this later.
- Need to brainstorm — How will we approach this in the
future? What do we do for next year? Will be recruiting
volunteers to help.
- The question for this afternoon: Does the community
feel like the capabilities we brought to the showfloor
works, and how we bridge this demo zone and the tech
program.
- Slide 4. There will be some questions: Scope? Target
participants?
- Scope - seems fairly trivial for this conference:
- • Optical photonics networks
- • Wireless - Optical integration
- • Monitoring & Measurement
- • QoS
- • Control plane - e.g., handling the ever increasing
cybersecurity complexity
- • Capacity / Capability - including low-level
programming of the network
- • Quantum - variety of demonstrations
- • AI & ML
- Target participants - what should the target be? Who
would be the participants in a workshop series, if we
start one?
- • Academia
- • National (and international) Laboratories, NSF
institutions
- • Industry R & D, labs
- • Startups coming out of hiding
- • Educators & Students (educating the next
generation, want them to participate in the future!)
- (notes: hyperscale cloud providers not specifically
mentioned. Later in the discussion, we did discuss
operators.)
- What would be incentives for the different communities
to participate in a workshop?
- Various measures of success: Publications, key notes,
invited talks, awards, citations, etc.
- A workshop would allow publishing - talks, posters,
and short papers.
- Also things like interesting student contents or
challenges, e.g., sponsored by industry and/or OFC
organizers. Categories might include supercomputing, big
data transfer on global-scale optical networks, etc.
- For awards, could ask for time on the Expo Theaters
showfloor schedule.
- Slide 5. More questions on the way forward.
- • Go for a workshop / symposium?
- • Half / full day?
- • How to have OFCnet demos optimally work with
preexisting Demo Zone (complement each other?)
- • Next year’s co-chairs? Go and no-go decision
points?
- Re: comparisons to SCinet & SuperComputing,
similar size conference. Academia, research labs, DOE
labs are very present there. Another incentive that
works there is that DOE and NSF program directors walk
around and have PIs report status on projects they have
funded.
- Slide 7. This is a sample of topics covered by a
workshop at SC known as Innovating the Network for
Data-Intensive Science (INDIS):
https://scinet.supercomputing.org/community/indis/
- Slide 8. Program for this afternoon.
- Time Title Presenter
- 0:00 Welcome, introduction Cees de
Laat
- 0:10 Introduction to OFCnet Marc
Lyonnais, chair OFCnet
- 0:20 OFC Demo Zone Marco Ruffini,
Ben Puttnam, DemoZone
- 0:30 Panel introduction Reza
Nejabati
- 0:35 Panel
- Andrew Lord, BT
- Hübel Hannes, AIT
- Richard Murray, Orcacomputing
- Daniel Kilper, TCD
- Inder Monga, ESnet
- Jörg-Peter Elbers, ADVA
- Each gets 5 minutes to present, 5 minutes to discuss
- 1:35 General discussion with audience on outcomes
& next steps
- Introduction to OFCnet by Marc Lyonnais
- OFCnet is a 3 year endeavor, this year being the
pilot, last year being “first light”
- Invited demonstrations: quantum network and
classical demonstrations, 19 total. Panel
discussions on the showfloor.
- Call for support from sponsors was very positive.
For 2024-2025, will also put out call for sponsors and
volunteers.
- Integrating OFCnet into demo session by Marco
Ruffini, Ben Puttman / OFC Demo Zone
- Demonstrations…reasons & motivations to use
OFCnet
- How can demos test the OFCnet environment in
advance, and what kind of expertise can be made
available on OFCnet
- Potential ideas for challenges (e.g., efficient
wavelength setup, adding paths over a given topology
while keeping change in OSNR below a given targeT).
Transmission over live fiber. Quantum
coexistence.
- Cees: several of the things mentioned are actually
things that the SCinet team offers
- Panel Introduction by Reza Negabati
- Panelists have ~5 minutes to present view
- Andrew Lord, BT
- OFCnet, very first impressions
- What would this be useful for, for me?
- - start-ups,
- - low impedance route for institutions to
demonstrate capabilities at OFC
- - being independently managed, adds some
credibility to demonstrated technology (e.g., not
necessarily just vendor “hype” or hand-waving)
- - a way for OFC to diversify into tangential areas
- What will be hard about this?
- - who manages it? Time-consuming, needs
funding/maintenance
- - how is the time/schedule managed? OFC
itself is short
- - how do you create a low entry point
- - how do you keep this going, how do you build
longevity and not run out of steam
- Inder: Reward mechanism. How do you reward the
people that come, so they either come back, or
consider other people to come?
- MarcL: One example was a quantum demonstrator, asked
to have OTDR data provided for the 3 fiber pairs, to
help them prepare and give them more confidence in
their demo.
- Andrew: Would like us to be a bit more creative,
e.g., not just putting equipment / kit on either ends
of a fiber that just happens to be in the
building. A stronger demo would consist of not
just be plugging a fiber in.
- Cees/Andrew: Managing resources, how to support a
large number of demonstrations all relying on the same
infrastructure?
- Hubel Hannes, AIT
- Coming from the quantum demo side.
- Feels that it is important to show that things (a)
can be racked/stacked, powered and it just runs (b)
coexists in a standard networking environment / rack
facility, (c) how it interfaces to existing networking
equipment.
- Bringing the equipment on-site and setting up the
demo prior to the conference does not present a big
challenge from their perspective. Can have a QKD
system up and running in a couple of hours given a
pair of fiber.
- Various examples of quantum/classical applications,
e.g., hospitals sharing genome data.
- What was really hard to do - show the benefits in
terms of security. Hacking competition??
- Would like to be able to show QKD in the context of
a larger network!
- What would this consortium benefit from OFC?
Possibility of bringing multiple vendors together,
maybe vendor interoperability / vendor agnostic
possibilities.
- Richard Murray, Orcacomputing
- OFCnet, very first impressions
- OFC Demonstration suggestion - Quantum Data Centre
of the Future
- Might view OFCnet as a way to:
- - get quantum technologies out of the lab, OFCnet
could be a good venue as they are thinking about
getting towards the datacenter
- - get quantum researchers out of the lab - expand
social networking aspects
- - e.g., talked about a previous demo day, got
datacenter people to “tell them everything that they
didn’t know” to help them understand how they might
get quantum technology into a datacenter environment
- Marc: Did you/your company see this activity as
worth it?
- - yes, not as much for science acceleration, but
definitely for the engineering aspect.
- Dan Kilper, TCD
- Using OFCnet to address the AI problem,
- Re: AI problem. Very few data sets available
(e.g., Microsoft dataset), largely due to privacy and
business issues for operators (e.g., NDA might prevent
them from releasing it - not necessarily holding back
for other reasons)
- NIST workshop on ML for Optical identified this
issue.
- Could we use the turn up & operation of OFCnet
to collect data sets? Make datasets and
reference results publicly available.
- Compare legacy (physics based) planning tools to ML
variations of these planning tools (e.g., could be an
annual competition, winners present algorithms and
what they did). Or live hack-a-thon sessions for this
years data set (e.g., give 24 hours/etc to work on the
data, announce the winners at the post-deadline paper
presentation)
- Example competition: AutoML Decathlon 2022
- Andrew Lord: Would one 17km pair of fiber be enough
to do anything interesting with?
- Marc: Could there be some kind of persistent
testbed, etc.
- Jorg-Peter Elbers: What is the benefit of generating
these datasets in the OFCnet context, vs. creating
this dataset somewhere else and still doing the
competition at OFC?
- Dan: One of the benefits of this venue is that it is
a neutral environment.
- Inder Monga, ESnet
- R&E perspective
- OFCnet motivation - we have a bunch of new/leading
systems being brought to OFC.
- What should OFCnet look like? More like a
datacenter, or a research lab?
- How can we leverage and/or showcase work being done
on existing testbeds, things that cannot be “shipped”
here? (e.g., NSF FABRIC, QuantNet)
- Jorg-Peter Elbers, ADVA
- One of the reasons why the demo zone exists is for
people to be able to see it and touch it.
- Demo Zone concept copied by ECOC. Researchers
love the demo zone, feels like we don’t want to change
that.
- There are hurdles in terms of shipping, logistics,
with organizing and putting on a demo.
- Would like to see a collaboration element, what
kinds of things can we do that would not otherwise be
possible? (e.g., vendors doing their own thing
inside their own ecosystem, but what about connecting
different people/groups together to do something
interesting?)
- Andrew/Marc/Jorg-Peter: Discuss about how
demonstrations
- Cees
- Some observations about the discussion. Liked
the idea of this being a neutral environment, it’s
okay to try to break it (if everything just works,
well you should just sell it). Bridge between
industry & tech programs. Bring the tech
program more to the showfloor (and vice-versa!).
- What is the next step?
- Have a good distribution in terms of participants
providing the experimental facilities (industry,
startup, academic, R&E / national laboratories)
- Did we miss targets?
- Dan: regarding students, I need buy-in from the
funding agencies that this is a worthwhile activity.
Program managers walking around the showfloor
expecting to see the results of their programs, need
NSF program directors (for example) walking around to
see these things.
- Ex-Adva now DTU: Did we miss the
operators? Surely they would want to play
too? Stress testing solution, is it intuitive
enough?
- Took a show of hands: who is from academia and/or
relies on publication to survive? (maybe
10-15%?)
- Marc: Focus Group that would investigate some of
these questions, we have a plan for 3 years, but how
can we address persistence / longevity (5 years, 10
years?)
- Inder/Cees: On workshops, some discussion about
coordinating with the technical program committee, is
there concern about overlap?
- https://scinet.supercomputing.org/community/indis/about
- Noted a lot of discussion about hack-a-thon and
challenges. Example of contests at SC which used (a)
supercomputer somewhere, live, (b) SCinet, and (c)
results in the theater/display on showfloor.
- Challenges can become quite interesting, especially
if you get some of the top groups in the world. It’s
about creating hype and generating interest.
- Marc’s observations, things that resonated from
this discussion:
- - what are the possibilities if we had access to a
“permanent” infrastructure?
- - value in what we can provide to the community /
researchers, whether it is data sets, etc
- - place to demonstrators to showcase new
technology while also providing a venue for those
demonstrators to learn from (e.g., quantum demo
example above)
- OFC2025 will be at a different venue, that in and of
itself will present its own challenge.
- Rodney Wilson: One of the key objectives is
promoting the transfer of information, it is a great
opportunity to bring together leading technologists
and researchers with the next generation.
- BOF closed 18h25.
- UvA participates with two demos in the OFCnet
booth 923
- UvA presence in OFCnet panel in EXPO III:
- UvA contributes two team members in OFCnet:
- Cees de Laat, team-lead OFCnet workshop
- JP Velders, team-lead OFCnet Security
|
