Joint Multilayer Planning of Survivable Elastic Optical Networks · 2016-05-24 · OFC 2016 Anaheim Convention Center, Anaheim, ... (Gb/ s) Reach (km ) Data slots Energy Consumption

Joint Multilayer Planning of Survivable Elastic Optical Networks

P. Papanikolaou, K. Christodoulopoulos, E. Varvarigos

Department of Computer Engineering and Informatics, University of Patras, Greece and Computer Technology Institute and Press – Diophantus, Patra, Greece

High Speed Communication Networks Laboratory

National Technical University of ATHENSOFC 2016

Anaheim Convention Center, Anaheim, California, US

N3: Network Architectures, Techno-Economics and Design TradeoffsM2K. Elastic Network Optimization

OFC 2016 (1/16)

MotivationTraditional IP over WDM network survivable approach:

IP layer is responsible for recovery (but may not be sufficient, unless dual plane approach is used)

Optical Protection is another approach (but wastes resources)

Dual Plane protection approach (reactive resilience exclusively on the IP layer):

Two network copies that mutually protect each other

Over-provisioning of IP interfaces and transponders (doubles equipment)

The dual plane approach was followed in the past to account for the lack of optical agility

Emerging optical technology is dynamic, enabling the use of Multi-layer Survivability Techniques:

Multilayer coordination allows more efficient resources usage in the network

Significant energy and cost savings potential

Network Model

OFC 2016 (2/16)

IP-over-Elastic Network

Planning an IP over an elastic network consists of 3 inter-related sub-problems:

IP routing (IPR)

Virtual Topology design

Routing of lightpaths and Modulation Level (RML)

Spectrum Allocation (SA)

Multilayer CAPEX model

Optical layer: flex-grid enabled OXCs and tunable - Bandwidth Variable

Transponders (BVTs)

Modular IP/MPLS router organized into 3 component classes: basic node

(3 types of chassis), line-cards, and short reach transceivers

Compared resiliency techniques

OFC 2016 (3/16)

Technique I (reference): Dual Plane

Reactive resilience exclusively on the IP layer:

Two network copies, each one dimensioned to carry 100% of the traffic

In case of failure the other network copy absorbs the total traffic of the network

Provides resiliency from optical link, optical node, and IP node failures

Technique II: Failure driven network design

Multilayer resilience on top of the dimensioned network (2 steps):

Step 1: joint multilayer dimensioning of the network for normal operation

Step 2: Examine the possible failure states and re-dimension both layers

Technique III: Integrated Multilayer (ML) survivable network design

Joint multilayer technique:

Dimension jointly both IP & optical layers, considering also all possible failure states of the

network

Proposed techniques II and III are used to recover from single optical link failures in this study

Dual Plane

OFC 2016 (4/16)

A and B variants of the network

Each network is dimensioned to

carry 100% of the traffic

Equal Cost Multipath protocol(ECMP) with 50:50 load sharingbefore failures

When a failure occurs, the entire

traffic is directed to the other

network

(IP/MPLS Fast-Reroute - FRR)

Response times below ∼50 ms

Failure driven network design (1/2)

OFC 2016 (5/16)

This resilience scheme consists of two steps:

1. Both IP & optical layers are dimensioned for normaloperation (i.e., no failures in the network)Objective of the design process:

min (capex, energy, spectrum)

2. The impact of all single optical link failures on IP links isaccounted and IP & optical layers are re-dimensionedaccording to the worst-case traffic

Higher response time than the dual plane approach

Some failures might require provisioning of lightpaths

High capacity efficiency: remaining capacity of primary can be used for backups additional backup capacity is shared among backup paths

of different connections in 1:∞ sharing but primary paths are fixed and are not jointly optimized

1:∞ failure driven network design (2/2)

OFC 2016 (6/16)

PrecalculationBoolean constants: Impact ofoptical link failures on IP links

ConstraintsBackup IP flows and the opticalcircuits are dimensionedaccording to the expected worst-case traffic for all single linkfailures

The backup path of aconnection can share with aprimary of another connection.

At both ends of the failedlightpath the same workingrouter interfaces are used.

ObjectiveThree-objective optimizationmin (CAPEX, Energy, Spectrum)

InputNetwork design withoutaccounting for failures

ILP_model

Integrated ML survivable network design (1/2)

OFC 2016 (7/16)

Holistic approach: jointly considers the cost of both layers(IP/MPLS and optical) and all possible failure states

Objective of the design process:

min (capex, energy, spectrum)

To survive from any single optical link failure extra capacity is added

Even higher capacity efficiency: remaining capacity of primary can be used for backups additional backup capacity is shared among backup paths of

different connections in 1:∞ sharing but primary and backup paths are jointly optimized (so better

than Technique II) Very complicated

Higher response time than the dual plane approach

Some failures might require provisioning of lightpaths

1:∞ Integrated ML survivable network design (2/2)

OFC 2016 (8/16)

ILP_model

Network Planning considering jointly:

IP/MPLS layer costs

Optical layer costs

All possible failure states

Constraints

Extra capacity is added to thebackup lightpaths of everyconnection

Backup lightpaths are link disjointto the primary ones

The backup path of a connection canshare with a primary of another

At both ends of the failed lightpaththe same working router interfacesare used

ObjectiveThree-objective Optimization

min (CAPEX, Energy, Spectrum)

Performance results

OFC 2016 (9/16)

DT topology

Traffic provided by operator (DTAG) for 2012

Plan the network from scratch for 2014-2024

Assumption: 35% uniform traffic increase per year

Spectrum slot: 12.5 GHz

Tunable transponder (BVT) – see table

Objective: weighted minimization of the cost, energy andspectrum, focusing on the first 2 (WC = WE =0.45, withthe remaining 0.1 to be the weight of the spectrum used)

Capacity (Gb/ s) Reach (km ) Data slot sEner gy Consumpt ion

(Wat ts)Capacity (Gb/ s) Reach (km ) Data slot s

Ener gy Consumpt ion

(Wat ts)

4000 5 183.6 3000 4 270

3000 4 183.6 2500 3 270

2500 3 183.6 1900 2 270

2200 6 432 750 9 432

1900 5 432 600 7 432

750 4 333 500 5 432

1.76cost of BVT (cost units)

40 100

Bandwidth Var iable Transponders

200 400

Multi-objective Optimization Pareto front

OFC 2016 (10/16)

Objectives: CAPEX minimization vs Energy minimization

Objectives: CAPEX minimization vs Spectrum minimizationThree-objective Optimization model

MINIMIZE (CAPEX, Energy, Spectrum)

Two-dimensional Pareto-optimal fronts(using: Integrated ML survivable network design model)

WSPECTRUM = 0.1 // WCAPEX [ 0.01 , 0.89] // WENERGY [ 0.01 , 0.89]

WENERGY = 0.1 // WCAPEX [ 0.01 , 0.89] // WSPECTRUM [ 0.01 , 0.89]

798

800

802

804

806

808

810

812

814

816

123000 123500 124000 124500 125000 125500 126000 126500 127000

minCAPEX

(co

st u

nit

s)

minEnergy (Watts)

DT traffic: 2018

450

455

460

465

470

475

480

485

490

495

500

200 250 300 350 400 450 500

minCAPEX

(co

st u

nit

s)

minSpectrum (GHz)

DT traffic : 2014

Illustrative Results (1/2)

OFC 2016 (11/16)

1 c.u.: the cost of a 100 Gb/s coherent optical transponder

Dual plane: exhibits the worst performance (copes with failures by over-provisioning)

The joint approach (Integrated ML survivable network design) overperforms the other two approaches,

(jointly dimensioning the network considering both layers’ costs and all the possible failure states)

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

2014 2016 2018 2020 2022 2024

Cap

Ex

(c.u

.)

Year

DT networkdual plane

failure driven design - O

Integrated ML survivable - O

Illustrative Results (2/2)

OFC 2016 (12/16)

Spectrum Utilization*

Dual plane double available spectrum better spectrum utilization(even though lambdas may be filled only up to 50% in error-free operation)

*for dual network, the utilization is for each network

Significant CAPEX savings, up to 46%, canbe achieved by considering optical linkfailures in joint ML network planning(techniques II and III)

0

5

10

15

20

25

30

35

40

45

50

2014 2016 2018 2020 2022 2024

Cap

Ex

Sav

ing

pote

ntial

(%

)

failure_driven integrated_ML

0

50

100

150

200

250

300

2014 2016 2018 2020 2022 2024

En

erg

y T

ran

sp

ort

ati

on

Eff

icie

ncy

(M

bs/J

ou

les)

Year

DT networkdual plane

failure driven design - O

Integrated ML survivable - O

Extended models (1/2) Work in progress

OFC 2016 (13/16)

The proposed joint resilience techniques:

exploit the agility of the optical layer

use more efficiently the resources in the network

yield significant cost savings

are unable to deal with optical node failures // IP layer failures

Extension of proposed models provides the same level of resilience with the dual plane approach:

optical nodes failures

full core router failures

Multilayer resilience techniques with two different level of failure analysis integration:

survive single optical link/node failures and core router failures

avoid overprovisioning the IP layer by exploiting the resources used for a primary path of one

connection and use them as a backup path for a node disjoint connection

Extended models (2/2) Work in progress

OFC 2016 (14/16)

The multi-layer resilience techniques overperform the single-layer dual plane approach, while achieving the same

level of resilience

Integrating failure analysis in the design process (Integrated ML survivable network design) provides an additional

gain, since it optimizes both primary and backup paths

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

2014 2016 2018 2020 2022 2024

Cap

Ex

(c.u

.)

Year

DT networkdual plane

failure driven design - O + IP

Integrated ML survivable - O + IP

Conclusions

OFC 2016 (15/16)

Single-layer Survivability mechanisms lead to over-provisioning of IP interfaces and transponders

Emerging optical technologies enables the use of Multi-layer Survivability Techniques, which:

allow more efficient resources usage in the network

provide significant cost and energy potential

Survivable multilayer network planning techniques

1:∞ Failure driven network design: on top of the dimensioned network

1:∞ Integrated Multilayer (ML) survivable network design

Both yield significant cost savings, with integrated being better but extremely complex

Survivability level offered:

Dual plane: optical link / optical node / IP node failure

Proposed ML survivability techniques: optical link failure

Extension of proposed models provides the same level of resilience with the dual planeapproach, achieving efficient resources usage and cost savings

Questions ?

OFC 2016 (16/16)