Andrew Wiedlea - Wireless FasterData and Distributed Open Compute Opportunities and (some) Use Cases
Wireless FasterData and Distributed Open
Compute
Opportunities and (some) Use Cases
Andrew Wiedlea
Science Engagement Team
9 February 2023
ESnet is developing our strategy to support “the wireless edge”
Motivation:
● Scientific progress will be completely
unconstrained by the physical location of
instruments, people, computational resources, or
data.
● Collaborations at every scale, in every domain,
will have the information and tools they need to
achieve maximum benefit from scientific facilities,
global networks, and emerging network
capabilities.
● ESnet will foster partnerships and pioneer the
technologies necessary to ensure that these
transformations occur.
A work in progress: a wide range of possibilities for what
services and capabilities may be selected
Science Use Case Drivers (short term)
● Convenience of IoT and the ability to deploy instruments
in non-laboratory settings and hostile or constrained
environments
● Within facilities, ease of relocation, equipment
movement, and management
● Explosion of wireless options, each with different
capabilities, cost models and uses makes data more fluid,
less expensive to transport
○ 5G, Private Wireless/CBRS
○ LoRa
○ Zigbee
○ Starlink, Orbcomm, Non Terrestrial Networks
○ mmWave
Longer term use case driver: Self Guided Field Laboratories and
an automation revolution for scientific activity
An enriched network + distributed compute + autonomous sensors = automated
scientific mass “production”
Abiven, Yves-Marie, Simon Bouvel, Thierry Bucaille, Leonard Chavas, Erik Elkaim,
Patrick Gourhant, Youssef Liatimi, et al. 2020. “Robotizing SOLEIL Beamlines to Improve
Experiments Automation.” In. JACoW Publishing, Geneva, Switzerland.
https://doi.org/10.18429/JACOW-ICALEPCS2019-MOPHA001.
BLUF: Enable the SGFL vision with PRP capabilities as part of Wireless
FasterData?
Can we deploy a ruggedized, affordable FIONA in the
field for edge compute services and satisfy power, PRP
mesh communications needs?
Leverage self-driving 5G work underway already by
PRP/NYU/ESnet Mariam Kiran, John Graham to test
these concepts in the field?
What opportunities are created for field science with a
PRP distributed 5G core?
We don’t know all the possible uses of wireless edge
and network distributed compute, and we can’t know
until we deploy something at reasonable scale….
And the reason is that our work both supports and
drives science possibilities.
SGFL Example: Water Cycle
APPROACH:
●Generating actionable information based on model
execution and data fusion
○ CSA: implementing model on HPC with autonomous data
ingestion
○ EESA: Executing the model to generate predictions
informed by lab (Bioepic) and field (East River)
measurements
○ CSA and EESA: integration of model with other large-scale
datasets to identify uncertainty distribution to guide
adaptation/feedback
●Integrated lab/field data streams to guide further
sampling in the field
o CSA: Data workflow and harmonization
o EESA: ET model sensitivity analysis targeting a range of field
scenarios
SGFL Example: NRT Earthquake Impact Analysis
Exploring a combination of NTN and cellular to
ensure rapid movement of structural response data
after a quake to inform large area damage simulation
based estimation
Petrone Floriana, Perez Rowland, Coates Jason, and McCallen David. 2023. “A
Biaxial Discrete Diode Position Sensor for Rapid Postevent Structural Damage
Assessment.” Journal of Structural Engineering 149 (3): 04022251.
SGFL Example: Urban & Radiation Mapping
Urban radiation effects are a complex combination
of:
● Soil and building materials
● Weather & atmosphere
● Human behaviors
Detection of anomalies is inherently a short range,
multi-mode/multi-int game, requiring both
distributed analytics and interesting patterns of
pushing data to the field / multi-level simulation
(1D-3D)
SGFL Example: Energy Grid
Work is underway to deploy a integrated
10K device/simulation/human in the loop
multi-lab platform for future energy grid
technology study.
Requires a low jitter/predictable latency
dedicated network, using ESnet.
Integrated predictable wireless (such as
via distributed core) and edge compute to
provide low delay data tagging and
reduction across standards may well
develop as a requirement.
SGFL Example: Offshore Wind Power
ECP’s ExaWind project/LBL/NREL research underway on
development of algorithms supporting wind turbine
design and placement optimization.
Very large, offshore tower concepts are being
developed. Larger is more efficient, but also will create
greater demands for more remote placement, increase
data movement/compute demands
Testbeds will absolutely provide opportunities for
wireless edge compute, present interesting
opportunities for research and control communications
(weather, tunable blades, fault and grid analysis)
Desired End State Thoughts
1) SGFL capabilities will be driven as much by what we as “an enriched network”
community dream up, as by what our current customers request
2) The integration of a federated wireless core with the enriched network could
jump start national progress towards SGFL
3) Now is a good time to take some risks; the technology space is fluid and we
can’t count on being right, but we also know that we can’t afford to fall behind
and let the market or near-peer competitors define our options
4) Let’s get nodes out into the field, and on orbit. I believe there could be an
interesting opportunity to work with one or more commercial providers*
Editor's Notes
Digital twin plus fluid data flow plus an enriched network (network+data+storage)