seminar at institut mines telecom may 3 “brighter networks: a · 9 © 2018 nokia nokia internal...
TRANSCRIPT
© 2018 Nokia1
Nokia internal use
Seminar at Institut Mines TelecomMay 3rd 2018
Sébastien Bigo
“Brighter networks: a glimpse into the future
of optical networks”S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia2
Nokia internal use
A vision of the future
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia3
Nokia internal use
The future of networks is all about time
S. Bigo, Seminar IMT, Paris - All right reserved
Timeto transport
The capacitychallenge
Time to queue
The latencychallenge
Time to reconfigure The automation
challenge
© 2018 Nokia4
Nokia internal use
Agenda
The capacitychallenge
The automation challenge
Space, the newfrontier
Q&A
The latency challenge
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia5
Nokia internal use
Spectrum[Hz]
Space
Spectral Efficiency [b/s/Hz]
≤
Bandwidth (B) beyond erbium limit
Approach non-linear Shannon Limit (SNR)
Expand fiber-space (M)
Smart Network Fabric : Capacity
C ~ B x M x log2(SNR)
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia6
Nokia internal use
High Baud Rates
High Order Modulation
10011010101
10100011101
10100111101
00101011010
10010100111
01001011001
Advanced Coding
Nyquist Filtering
Smart Network Fabric: Solving for SNR with PCS
Probabilistic Constellation
Shaping
The final piece of the super coherent puzzle
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia7
Nokia internal use
Building a new langage, easier to decipherBased on existing symbols
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia8
Nokia internal use
Discrete QAM formats perform well only at specific distances
Distance
Capacity
Shannon Limit approaching
optimal capacity at any distance
25%Shaping gain
PCS
PCS
PCS
Smart Network Fabric: Solving for SNR with PCS
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia9
Nokia internal use
Smart Network Fabric: Solving for B
-50
-45
-40
-35
1480 1510 1540 1570 1600 1630 1660
Wavelength (nm)
Po
we
r (d
Bm
)
EDFA
SOA
100nm
C-Band L-BandS-Band
250 channels, each carrying >400 Gb/s, making the first 100-nm continuous band transmission of 115 Tb/s over 100 km (reported at ECOC 2017)
375
400
425
450
475
500
1500 1520 1540 1560 1580 1600
Ch
. n
et
bit r
ate
(G
b/s
)
Wavelength (nm)
C band L bandS band
Total throughput: 115.9 Tb/s115.9Tbit/s
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia10
Nokia internal use
Smart Network Fabric: Solving for M
Single fiber, more cores Single fiber, more modes
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia11
Nokia internal use
Smart Network Fabric: 1Pb/s System in 2024 ?
~ 1Tb/s
Each cell = Single mode with maximum spectralefficiency (PCS inside)
6 to
40
sp
atia
l p
ath
s
50 (C) to 100 cells (C + L)
Spectrum(more wavelengths)
Space(more modes, more fibers)X100
Shannon (more PCS)
~ 1Pb/s
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia12
Nokia internal use
Agenda
The capacitychallenge
The automation challenge
Space,the new frontier
Q&A
The latency challenge
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia13
Nokia internal use
MICROSFT cloud networkS. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia14
Nokia internal use
© Nokia 2016
Telecommunicationproviders
Prescribers in optics
Followers in IP/Ethernet
Over the topcontent providers
Prescribers in IP/Ethernet
Followers in optics
Two types of players competing to define and rule the 2025 global cloudIP + Optics
Te
lco
s
OT
Ts
© 2018 Nokia15
Nokia internal use
Two types of players, two competing visionsCreating the future with the best of both
Try and learnContinuously adjustAlways zero marginsWorking with just-enough performance
Design and optimizeSet and forget
Guaranteed (large) marginsSeeking top performance (bit-rate/distance)
Detect and mitigateAutomatic troubleshoot
Seek for maximum quietness(Zzzzz…. $$$$)
S. Bigo, Seminar IMT, Paris - All right reserved
Optics today(Telco centric)
Optics tomorrow
Quotes from Najam Ahmed, Facebook, OFC 2016
© 2018 Nokia16
Nokia internal use
Two types of players, two competing visionsCreating the future with the best of both
Push down demands to the networkProvision and monitorAll programmable, very dynamic network If equipment fails, remove from network,Perform troubleshooting later
Plan resources well in advance (e.g. 15y ahead)
secure them before establishing servicethen allocate demands.
Optics today(Telco centric)
Optics tomorrow
Quotes from Najam Ahmed, Facebook, OFC 2016S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia17
Nokia internal use
Two types of players, two competing visionsCreating the future with the best of both
Connection-based approachReserved, even if partially filledGuaranteed performance and jitter-free
All-packet approachFully flexible, can be shared
Bandwidth and latency depend on load
Optics today(OTT centric)
Optics tomorrow
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia18
Nokia internal use
Two types of players, two competing visionsCreating the future with the best of both
a network to iterate, not architecta network which fails fast instead of a fail-proof onea network which pushes down demands to the network, not planned 15y aheada network still with guaranteed delivery, but only when needed
Optics tomorrow
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia19
Nokia internal use
Living networks, accounting for timeBrighter optical networks
..
Flexible TRX
SDNControl
New demand
Resourceallocation
New hardware?
Predict
Measure
Demands (t)Learn from the pastPredict the future
Hardware resources (t)Learn from the pastPredict the future
f(t)+
-
From set-and-forget to try-and-learn
Margin
predictor
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia20
Nokia internal use
Marginless live network planning with machine learning
Network margins (as rephrased by Augé et al., OFC’13):
Unallocated margins – when the achievable distance exceeds the transmission distance
Design margin – unwanted, worst case QoT prediction (imperfect model + input uncertainties)
System margins – time-dependent, fixed margins (e.g. polarization effects, ageing, network load)
Many recent works on QoT prediction with learning
Oda et al., JLT, vol. 35, no. 8, 2017
Sève et al., OFC’17
Yan et al., ECOC’17 PDP
Bouda et al., JOCN vol.10 (1), 2018
…
[Sève et al., OFC’17]S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia21
Nokia internal use
• Key goal
Fibers type for rent (especially dark fibers) are sometimes either unknown or mistaken for another type, hence require worst-case design
• Key disruption
Automatically detect fiber type without traffic interruption by monitoring and correlating the accumulated chromatic dispersions extracted from coherent receivers of established lightpaths.
S. Bigo, Seminar IMT, Paris - All right reserved
Automated fiber type identification for cost reduction without traffic interruptionGoal and disruption
Reduction of network cost thanks to a better knowledge of type of fiber
Number of regenerators: 5Fiber 1 (SMF): high perf.
Fiber 2 (LEAF): low perf.
Unknown LEAF
Worst case scenario :
Number of demands: 0 With monitoring
Number of regenerators: 3
?
??
Number of demands: 1Number of demands: 2Number of demands: 3Number of demands: 4
R$$$$$R$$$
Unknown fiber type
SDN platform
vFiber
Type vCost Paris
London
Lyon
Strasbourg
Amsterdam
Bonn
Hamburg
Zurich
Brussels
Number of regenerators: 4
R$$$$
© 2018 Nokia22
Nokia internal use
Reduce overprovisioning (margins) on new lightpaths by reducing the uncertainties of the network characteristics thanks to
(1) monitoring
(2) learn from previously established lightpaths
E. Sève et al., OFC’17
Quality of transmission model,
Application to power uncertainties
Example – Containing design marginsMarginless live network planning with machine learning
No learning
# Lightpaths
SN
R e
rro
r (d
B)
0
-1
-2
1
2
3
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia23
Nokia internal use
Track mechanical stress thanks to real-time State of Polarization monitoring
SOP Monitor embedded in coherent receiver
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia24
Nokia internal use
Use fiber as a sensor to avoid outage while providing root cause analysis of incidents
Track mechanical stress thanks to real-time State of Polarization monitoring
Embedded in a Coherent Receiver
Brings enhanced security and resilience to the fiber infrastructure by - detecting fiber break - classifing event with machine learning
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia25
Nokia internal use
Agenda
The capacitychallenge
The automation challenge
Space, the newfrontier
Q&A
The latency challenge
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia26
Nokia internal use
When cloud contaminates transportDistributed (edge) cloud networking
New stringent requirements, new traffic patterns
Beyond industry 4.0, on-line gaming, VR/AR, and 5G (cloudRAN)
- Standard packet-based networks fail because of latency
fluctuations (jitter)
- Circuit-based networks are not economically viable because too rigid/costly for the volatility of DC-like traffic
Need for a single convergent platform
- to host most time-sensitive apps, where lower and higher bounds should match (deterministic)
- reactive to the traffic pattern of a DC (dynamic)
In today’s DC(measured)
What we needfor time sensitive
applications
Fre
qu
en
cy
Fre
qu
en
cy
Latency (µs)
Latency (µs)Dynamic Deterministic Networks
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia27
Nokia internal use
Making the best of two opposing strategiesContaining latency in 5G networks
packet
switching
circuit
switching
Bell Labs proposal:
Optical Ethernet (OE) between edge DCs
Cloud Burst optical slot switching (CBOSS) inside DCs
Natively designed for best effort traffic
Effortless on-the-flyutilization of resources
Increasing latency with increasing load
Decreasing throughput with increasing load
Statistical multiplexing of data
Natively designed for business traffic
In-advance reservation of resources
Deterministic, guaranteedlatency
Constant, guaranteed throughput
Secure separation of data flows
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia28
Nokia internal use
Your worst waiting experience ? .
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia29
Nokia internal use
Bell Labs approach : get a time, why wait in line ? Distributed Dynamic Networking becoming real
Eliminate latency fluctuations by
- introducing (electrical or optical) time slots of fixed duration as data containers
- reserving slots for time-sensitive traffic, opportunistic queueing for best effort traffic.
- Automated orchestration from one domainto the next for strict end-to-end latencycontrol and L1 slicing
* across access fronthaul
* backhaul across DC and metro.
Change the user’s waiting experience as Disneyland did it with Fastpass
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia30
Nokia internal use
Agenda
The capacitychallenge
The automation challenge
Space, the newfrontier
Q&A
The latency challenge
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia31
Nokia internal use
High capacity optical satellite communicationSpace, the new frontier for coherent systems?
Optical feeder link
Optical link
Internet Ground station
Optical link
Clients
Countries with poor Internet connection diversity
Landlocked countries
Risk of disconnection:
SevereModerateLowRobust
RF link
Key goal (What we are trying to achieve):
Bring internet services to populations in countries in isolated or landlocked world regions and enhance reliability of existing internet services
Key disruptions (How we are thinking differently):
Leverage Nokia’s leadership in coherent transmissions to develop:
space-proof hybrid RF/optical satellite repeaters, with 10X greater Gbit/s/Watt
the corresponding optical ground stations, integrated with terrestrial networks
S. Bigo, Seminar IMT, Paris - All right reserved
© 2018 Nokia32
Nokia internal use
Agenda
The capacitychallenge
The automation challenge
Space, the newfrontier
Q&A
The latency challenge
S. Bigo, Seminar IMT, Paris - All right reserved
Nokia internal use