Upload
lyhuong
View
223
Download
1
Embed Size (px)
Citation preview
1 © Nokia 2015
Bell Labs
Design of Low-Margin Optical Networks OFC 2016 – Paper Tu3F.5 (invited)
• Dr. Yvan Pointurier, department head, Bell Labs
• 22-03-2016
Public
2 © Nokia 2015
1. Margins
2. Reduction of system margins
3. Reduction of unallocated margins
4. Reduction of design margins
5. Conclusions
Public
Agenda
Presentation funded by the European Commission through the H2020 ORCHESTRA project
http://orchestraproject.eu/ ICCS / NTUA
3 © Nokia 2015
• Margins
- Trade cost for reliability
• “margin of safety” = failure load
design load -1
• buildings: 100%; cars: 200%; planes: 20-200%
- (Limited) margins are wanted by customers
• Design of optical networks
- Green-field: given nodes, links, traffic demand allocate resources (equipment: optical transponders, types of transponders, regenerators, IP ports) to minimize some cost metric (CAPEX)
- Brown-field: allocate new traffic demands to minimize cost metric
• How to decrease margins when designing a network?
Public
Margins and design of optical networks
By Arriva436 (Own work) [GFDL or CC BY 3.0, via Wikimedia Commons]
4 © Nokia 2015
• [? dB] Operator margin
• [0.4 dB] Fast time-varying penalties
- Typically: polarization effects
- Use worst-case
• Slow time-varying penalties
- [1.5-3 dB] Varying network load varying nonlinear effects
- [several dB] “Ageing”:
• [0.7 dB] Amplifier noise figure (NF)
• [1.6e-3 dB/km/year] Additional losses due to splices after fiber cuts
• [0.05 dB/filter] Filter misalignment due to laser detuning
• [0.5 dB] Transponder
Public
Margins System margin
Beginning of Life - BoL End of Life - EoL Planning
time
Signal “quality” (SNR, Q2 factor, reach …)
FEC
limit
Planned
BoL
value
System
margin
Fast time varying penalties
Ageing
Network
loading
(Nonlinearities)
Operator margin
Most data from: Augé (Orange), OFC 2013, OTu2A.1 Pesic et al. (Bell Labs), OFC 2016, M3K.2
5 © Nokia 2015
Unallocated margin
=
theoretical perf. of equipment
-
perf. really needed to satisfy demand
• Need reach (SNR) = x km (y dB) for demand of z Gb/s but equipment can provide x’>x km (y’>y dB) SNR
Unallocated margin (dB) = y’-y
• Driven by discrete granularity of equipment performance
Public
Margins Unallocated margin
Beginning of Life - BoL End of Life - EoL Planning
time
Signal “quality” (SNR, Q2 factor, reach …)
FEC
limit
Planned
BoL
value
System
margin
Planned
EoL
value
Unallocated
margin
Fast time varying penalties
Ageing
Network
loading
(Nonlinearities)
Operator margin
6 © Nokia 2015
Design margin
=
real performance on the field
-
planned performance
• Sources:
- Inaccuracy of the physical layer models underlying the Quality of Transmission (QoT) tool
- Inaccuracy of the inputs of the Quality of Transmission (QoT) tool
- Estimation: <2 dB
Public
Margins Design margin
Beginning of Life - BoL End of Life - EoL Planning
time
Signal “quality” (SNR, Q2 factor, reach …)
FEC
limit
Actual signal
quality
Planned
BoL
value
System
margin
Planned
EoL
value
Unallocated
margin
Fast time varying penalties
Design margin
Ageing
Network
loading
(Nonlinearities)
Operator margin
7 © Nokia 2015
• Consider the 600 km path below with a 100G PDM-QPSK lightpath
- System:
• BoL: 4.7 dB (0.4 dB fast varying penalties, 2.3 dB slow ageing, 2dB nonlinearities )
• EoL: 0.4 dB (fast varying penalties)
- Unallocated: 6dB (unloaded reach=7100 km)
- Design: 1 dB (assumed)
- Total: 11.7 dB @ BoL, 7.4 dB @ EoL
Public
Margins An example
Beginning of Life - BoL End of Life - EoL Planning
time
Signal “quality” (SNR, Q2 factor, reach …)
FEC
limit
Actual signal
quality
System
margin
4.7 dB
Actual
margin
at EoL:
7.4dB
Actual amplitude
of the fast varying
penalties
Unallocated
margin
6 dB
Fast time varying penalties
Design margin
1 dB
Ageing
Network
loading
(Nonlinearities)
Operator margin
TX RX 100km
node
NF=4.5dB
Actual
margin
at BoL:
11.4dB
System
margin
0.4 dB
8 © Nokia 2015
1. Margins
2. Reduction of system margins
3. Reduction of unallocated margins
4. Reduction of design margins
5. Conclusions
Public
Agenda
9 © Nokia 2015 Public
System margin Load
700 km 6000 km 32000 km D.J. Ives et al., PNC, 29 (3), 2015; Bononi et al., NOC 2014.
+25% +50% +300%
• Target:
- network capacity (green field)
- decreased or delayed investment; network life
• Method: power, spectrum, rate allocation
• Gain: +25-300% capacity (depending on network diameter)
• Requirement: flex-grid and rate TRX/ROADMs; control plane.
• Possible improvement: flex-grid, multi-layer
• Challenges: very fine (per-lightpath) power management; requires network re-optimization during operation
10 © Nokia 2015
• Target:
- decreased or delayed investment
- increase network capacity during first years of operation
• Method: routing, spectrum, mod. multi-period allocation
- change modulation + deploy new equipment as network ages and traffic increases
- lightpath datarate may change leverage unallocated margin
• Sample study:
- System margins around 2 dB (ageing)
- Nonlinear effects: worst case – fully loaded links
• Gain: -10% TRX cost during first few years – few % at EoL
• Requirement: control plane, flex-grid and rate TRX/ROADMs
• Possible improvement: flex-grid, multi-layer, account for network load (nonlinear effects)
• Challenges: change signal rate during network operation
Public
System & unallocated margins Ageing & unallocated
See also Dupas et al., OFC 2016, Th3I.1 – Hitless 100 Gbit/s OTN Bandwidth Variable Transmitter for SDN
Pesic et al, OFC 2016, M3K.2
Traffic increase = 20% / year
0
2
4
6
8
10
12
14
16
0 1 2 3 4 5 6 7 8 9 10
Co
mp
ou
nd
co
st
sa
vin
g (
%)
Year
Cost erosion = 20% / year
Cost erosion = 10% / year
Cost erosion = 5% / year
No cost erosion
11 © Nokia 2015
1. Margins
2. Reduction of system margins
3. Reduction of unallocated margins
4. Reduction of design margins
5. Conclusions
Public
Agenda
12 © Nokia 2015
• Target: network capacity , network cost
• Method: (multilayer) rate allocation
• Gain: TBD
- Note: previous multi-year study also leveraged unallocated margins
• Requirement: multilayer control plane, flex-rate TRX
• Challenges: optical TRX at high rate deployed at commissioning; online re-allocation
Public
Unallocated margin Traffic growth
IP router
flex 100G
flex 100G
600 km
flex 200G
IP router
flex 200G
13 © Nokia 2015 Public
Unallocated margin Protection
Gold 100G
Best effort 100G
IP router
100G best effort traffic + backup path for gold traffic
flex 100G
IP router
Gold 100G
Best effort 100G
FEC + operator margin + fast time-varying system margin
unallocated and slow-varying system margin
flex 100G
• Target:
- (brown field) network capacity
- protection with differentiated classes of service
• Method: rate allocation
• Gain: TBD
• Requirement: control plane, IP/optical cooperation, flex-rate TRX
• Challenges: willingness to differentiate classes of service
flex 100G
flex 100G
100G gold traffic (primary path)
QoT
time failure
QoT
time failure
FEC + operator margin + fast time-varying system margin
Unalloc. and slow-varying sys. margin
14 © Nokia 2015 Public
Unallocated margin Protection
Gold 100G
Best effort 100G
IP router
flex 150G
IP router
Gold 100G
Best effort 100G
flex 100G
flex 150G
• Target:
- (brown field) network capacity
- protection with differentiated classes of service
• Method: rate allocation
• Gain: TBD
• Requirement: control plane, IP/optical cooperation, flex-rate TRX
• Challenges: willingness to differentiate classes of service
FEC + operator margin + fast time-varying system margin
unallocated and slow-varying system margin
QoT
time
100G best effort traffic + backup path for gold traffic
100G gold traffic (primary path)
QoT
time failure
failure
flex 100G
15 © Nokia 2015
1. Margins
2. Reduction of system margins
3. Reduction of unallocated margins
4. Reduction of design margins
5. Conclusions
Public
Agenda
16 © Nokia 2015
• Uncertainty on Quality of Transmission models (assuming perfect inputs)
- Ongoing work in many teams
- Always include more effects, more inputs: per-channel power, interactions of channels with different modulation formats, etc.
- More inputs more knowledge of physical layer is needed
- Some inputs can be set or measured and considered as known (“perfect”) at network deployment; for the other inputs…
• Uncertainty on Quality of Transmission inputs
- Some inputs are not or cannot be known prior to network deployment
• Sometimes even seemingly straightforward inputs (fiber type or length, dispersion map …) are not known!
• Exact characteristics of network equipment cannot be known prior to deployment
- On-the-field measurements (monitoring) only helps in brown-field scenarios
Public
Design margin Monitoring
17 © Nokia 2015
• Target:
- (brown field) network capacity , network cost
- lightpath establishment, network re-optimization
• Method: spectrum allocation
• Gain: +10-120% regenerators (10G OOK network)
• Possible improvement: flex-grid, multi-layer, multi-rate
• Requirement: knowledge of the network
• Challenges: margin estimation, probabilistic design
Public
Design margin QoT inputs
Reference: QoT > QoTFEC + m0 (no uncertainty) Proposed: QoT > QoTFEC + β.m(ΔP, lightpath)
Zami et al., BLTJ 14 (4), 2010
Uncertainty on the channel power (ΔP in dB)
0
50
100
150
0 0.5 1
% a
dd
itio
na
l re
ge
ne
rato
rs
0
50
100
150
0 0.5 1
β = 1 (84% certainty)
β = 2 (97.5% certainty)
Margins
18 © Nokia 2015 Public
Design margin: Monitoring Training
Planning (resource
allocation, QoT prediction)
Equipment: E0+ε0
Demands
E1+ε1 Physical layer parameters estimator
Monitors
E1+ε2
Design margin: m=f(ε0)
(ε2<ε1)
Planning (resource allocation, QoT
prediction)
E1+ε2
New demand
E3+ε3
Lower design margin: m'(ε2) <m
Prediction
Refined QoT predictor inputs
More accurate inputs Less over-
dimensioning
Estimation framework e.g. regression, machine learning
19 © Nokia 2015
• Target: (brown field)
- network capacity , network cost
• Method: monitoring, estimation framework, spectrum allocation on lightpath establishment and network re-optimization
• Gain: up to #regenerators / 2
• Possible improvement: flex-grid, multi-layer, multi-rate
• Requirement: monitoring, control plane
• Challenges: margin estimation (accurate monitoring), probabilistic design, online re-allocation
Public
Design margin QoT inputs
Sartzetakis et al,
OFC 2016, Tu3F.2
Q2
Q1
Q3?
Fully loaded links Estimated QoT inputs
Perfect knowledge of QoT inputs
To
tal n
um
be
r o
f re
ge
ne
rato
rs
Network load (Erlang)
20 © Nokia 2015
• Different types of margins different impacts
- Commissioning (green-field) network cost
- Upgrade (brown-field) network cost
- Longer network life
• Addition of small gains to lead to substantial overall gains
• Most key building blocks are available (or close)
- Flex-everything, (most) monitors, control plane
• Not a free lunch
- per lightpath power settings
- dynamics / online re-optimization
Public
Conclusions
Thanks to: T. Zami, J. Pesic, P. Ramantanis, P. Jennevé, N. Rossi, C. Delezoide and S. Bigo + ORCHESTRA project members.
Total cost of ownership
- (more) monitors
- (more) complex control plane
Bell Labs recruits!
Researcher positions available. Contact: [email protected]
22 © Nokia 2015
Public
Copyright and confidentiality
The contents of this document are proprietary and confidential property of Nokia. This document is provided subject to confidentiality obligations of the applicable agreement(s). This document is intended for use of Nokia’s customers and collaborators only for the purpose for which this document is submitted by Nokia. No part of this document may be reproduced or made available to the public or to any third party in any form or means without the prior written permission of Nokia. This document is to be used by properly trained professional personnel. Any use of the contents in this document is limited strictly to the use(s) specifically created in the applicable agreement(s) under which the document is submitted. The user of this document may voluntarily provide suggestions, comments or other feedback to Nokia in respect of the contents of this document ("Feedback").
Such Feedback may be used in Nokia products and related specifications or other documentation. Accordingly, if the user of this document gives Nokia Feedback on the contents of this document, Nokia may freely use, disclose, reproduce, license, distribute and otherwise commercialize the feedback in any Nokia product, technology, service, specification or other documentation. Nokia operates a policy of ongoing development. Nokia reserves the right to make changes and improvements to any of the products and/or services described in this document or withdraw this document at any time without prior notice. The contents of this document are provided "as is". Except as required by applicable law, no warranties of any kind, either express or implied, including, but not limited to, the implied
warranties of merchantability and fitness for a particular purpose, are made in relation to the accuracy, reliability or contents of this document. NOKIA SHALL NOT BE RESPONSIBLE IN ANY EVENT FOR ERRORS IN THIS DOCUMENT or for any loss of data or income or any special, incidental, consequential, indirect or direct damages howsoever caused, that might arise from the use of this document or any contents of this document. This document and the product(s) it describes are protected by copyright according to the applicable laws. Nokia is a registered trademark of Nokia Corporation. Other product and company names mentioned herein may be trademarks or trade names of their respective owners.