researchchallenges formulticastin carriernetworks · departamento de ingeniería telemática mpls 0...

MPLS 0Departamento de Ingeniería Telemática

Research Challengesfor Multicast in Carrier NetworksDavid LarrabeitiU. Carlos III de Madrid


Elastic Networks thematic network

New Elastic Networking Paradigms for a world radically-based on Cloud and Fog Computing

D. Larrabeiti, J.A Hernández


Objetives◆ Objetive: Strongly Cloud-Fog-

Oriented Networking

◆ Elastic Network Concept: ● Research towards the architecting a

more agile, flexible, programmable network with ultra-low deployment & OAM costs and energy efficient.

● Putting ultra-low latency and ultra-high capacity where needed

◆ Open to collaboration with interested third parties● Phd students exchange, joint papers,

joint thesis


Topics● Addressed Topics

§ EON

§ 5G

§ Fronthaul/backhaul transport: SDM, Ethernet, …

§ Power over Fiber§ Control plane SDN/NFV

§ Fog2Cloud computing

§ Opportunistic network/computing

§ Advanced forwarding techniques§ Optical Multicast

§ ..

Elastic Networks

Network

UC3M Larrabeiti David

CTTC Martinez Ricardo

UAN Reviriego Pedro

UPC-CRAAX Masip Xavi

UPC-CCABA Solé Josep

UPCT Pavón Pablo

UPV Casares Vicente

UdG Calle Eusebi

UVA Durán RamónJ.

UPM Fernández David

TelefónicaI+D López Víctor


2nd workshop◆ OPEN Free open workshop : ElasticNETS 2017

● 2nd International Workshop on Elastic Networks Design and Optimisation (ELASTICNETS 2017)

● June 13-14, 2017Vilanova, Spain

● https://elasticnetworks.org/2nd-workshop● Dates: send paper or register before June 1st

§ No proceedings, informal discussion of on-going work

§ pre-located to

Elastic Networks


Research Challengesfor Multicast in Carrier NetworksDavid LarrabeitiU. Carlos III de Madrid


WHY OPTICAL MULTICAST ?


CEHDTV decoder

1:M

1:C

1:R

CE

PE Router/switch

ROADM orOptical switch

Aggregation

Core

Tree-rootCore

….

….…. ….

….

HDTV

Metro

Access

APP1: TV


App2: IP mcast: eventually a zombie App3: VPLS

◆ Problem: VPN LSP trees do not match● Penalty: tail-end overhead

PE’s

•Per-VPN Forwarding entries in the core

• Only affordable for a well selected set of high speed multicast sessions

•VPN FEC awareness in the core !!!

•If Optical: a light tree per VPLS• NOT VIABLE:• Consequence: 100 papers on aggregation strategies


How aggregation converts your multicast light-tree into a broadcast tree

Huge group (50% of 1000)Large groups (10% of 1000)

Small groups (1% of 1000)


25 years of Optical Multicast Research

◆ 90% : Resource allocation: ILP+heur● Light-tree RWA ● Multicast Traffic Grooming over Light-trees

§ NO SENSE IN PRACTICE FOR SPARSE SCENARIOS !!!

◆ 10%: Switch Architectures● Broadcast-and-select● Reconfigurable Splitters● Tap-and-Continue● Sharing of MC unit● Tap-Configurable-2-Split-and-Continue (T2SC)


Is the Research on Multicast Terminated?


Challenge: adapt to App3 ?

◆ OCS via:● Use a broadcast channel● Engineer protocols to use it● EON adaptation to demand

◆ Alternatives to OCS? KEYS:● Sub-lambda switching● Source routing (SPRING/BIER)● Long-haul burst-mode transmission


Approaches to multicast◆ Circuit-Switched Light-tree (CSLT)

● High overhead => Aggregation of groups

◆ Traffic grooming in CSLT● Electronic grooming● High overhead if small groups, and/or few light-trees

◆ OBS light-tree● High loss: mcast => higher probability of collision

◆ Optical Packet Switching● Mature enough ?

◆ Optical light-trail● Electronic switching & copying among trails● No overhead, but suboptimal

◆ Optical light-tree trail


a) Tap & Continue

c) Tap & Two Split

In 1 Out 1Out 2

In 1 Out 1Out 2

In 1 Out 1Out 2

TxRx Tx

Rx Tx

b) Tap & Continue & Add

In 1 Out 1Out 2 In 1 Out 1

Out 2Rx TxRx

d) Tap & Erase

In 1 Out 1Out 2

Rx Tx

e) Tap & Add

In 1 Out 1Out 2

f) Tap & Two Add

In 1 Out 1Out 2

Rx Tx

Rx

Tx

Rx Tx

Rx Tx

Fig. 2. T2Sg Switch functionalities

T2Sg Switch Functionality


Implementation

Add/Drop Switch

3dBCoupler

3dBCoupler

PBAR

Dn, a, LPin

P1

Delay Line

Pout(2)Detector

Data for Node

Ptap

Dn

Dn

Optical Fiber Integrated Optics

PCROSS

Data + Label

Data + Label

TS

Label Interpreter

Pout(1)

Tap

Fig. 3. Structure of the proposed Tap 2-Split Traffic Grooming Switch (T2Sg).


Fig. 1. 4x4 2-STCg OXC node structure.

Full WDM switch


T2Sg: sample multicast requests & light-tree

A

BX

Z

Y

T

A è{X,Y,Z,T}

V

W

B è{C,D} , B è{Z,V} , B è{Z,T,W,X}

C

D

Light Tree 1Light Tree 2


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50

60

70

80

90

100Transmission Saving on T2Sg Broadcast tree w.r.t. Static T2S Broadcast tree for L=4, L=6 and L=12

p = probability of node membership to multicast group

% b

andw

idth

sav

ing

w.r.

t. br

oadc

ast

L=4L=6L=12

Fig. 12. Relative bandwidth saving featured by T2Sg with respect to T2S in a shared broadcast tree

Relative bandwidth saving

Huge group

Large group

Small groupSavingw.r.t.OCS light-tree


Path attenuation

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 17

8

9

10

11

12

13

14

15Average Attenuation due to Insertion Loss in the path for depth L=4

p = probability of node membership to multicast group

Atte

nuat

ion

(dB)

Multicast over T2Sg broadcast treeMulticast over T2S broadcast tree

Fig. 13. Average attenuation due to insertion loss in a path root-leaf for

a 4-node-deep binary light-tree

dBHuge groupLarge group

Small group

StaticBroadcast tree

Sub-lambda swBroadcast tree


Conclusions/Challenges◆ A lot of existing OCS multicast papers but..◆ NO effective solutions for sparse multicast◆ Little effort devoted to map electronic advances

to optical layer● BIER/BF header optical processing● Multicast Segment Routing ?

§ .. into optical● Undone: Leverage Broadcast Channel by L3/L2

protocols● Multicast OPS● Light-tree trail multicasting ?


• Acknowledgements:

Elastic NetworksTEXEOONDM TPCTibor Cinkler

Thanks!

researchchallenges formulticastin carriernetworks · departamento de ingeniería telemática mpls 0...

Documents