Optical Brust Switching network simulatin omnet

An OMNeT++ Model for the Evaluation of OBS Routing Strategies

A.L. Barradas, M.C.R. MedeirosCenter for Electronic, Optoelectronic and Telecommunications (CEOT)

Faculty of Science and Technology, University of Algarve, Portugale-mails: {abarra, cmedeiro}@ualg.pt

Presentation for

OMNeT++ 2008

SIMUTOOLS 2008March 7, 2008, Marseille, France

mailto:cmedeiro%[email protected]

Outline

● Introduction● Proposed Routing Strategies● Network Simulation Model● Performance Results● Conclusion

Alvaro Barradas for OMNeT++ 2008, March, 7, Marseille, France 2 / 24

Introduction

● Optical Switching Paradigms– Optical Circuit Switching (OCS)

– Optical Packet Switching (OPS)➔ Optical Burst Switching (OBS)


IntroductionOptical Switching Paradigms

● OCS– Establishes an end-to-end pipe

(wavelength/lightpath)

✔ Mature technology

✔ Good QoS for admitted traffic

✗ Large wavelength granularity

✗ Long reconfiguration delay

✗ Not suitable for bursts or short-lived sessions



● OPS– Similar to e-packet switching

✔ Familiar architecture

✔ Great flexibility and statistical multiplexing

✗ Per packet processing overhead

✗ Stringent synch & switching requirements

✗ Technological limitation ( ... no optical RAM!)



➔ OBS– “Leverages the best of electronics and optics ” *

✔ Moderate processing overhead

✔ Asynchronous switching

✔ Low signaling delay and high statistical multiplexing

✗ Lack of optical RAM causes burst dropping

✗ Moderate/Fast switching also needed

✗ Relatively new, unfamiliar technology

* Chunming QiaoInt’l Workshop on Future Optical Networks,March 5, 2006, Anaheim, CA, USA.



A comparison table by:K.C. Chua, M. Gurusamy, Y. Liu and M.H. PhungQuality of Service in Optical Burst Switched NetworksSpringer, 2007


IntroductionOBS reliability issues

“Leverages the best of electronics and optics”, but ...

Burst loss:

● One way reservation● Burst transmission without waiting for permission● demands > resources

( Collisions )Contentionresolutionscheme

Space

Time

Wavelength

Burst dropAlvaro Barradas for OMNeT++ 2008, March, 7, Marseille, France 8 / 24

Proposed Routing StrategiesMinimize Maximum Congested Link (MCL)

Minimize ³MAX (1)

Subject to

X

v2Vs;d¾v = 1; 8s; d 2 N (2)

X

s;d

X

v2Vs;d¾v £ pvl £ ts;d · ³MAX ; 8l 2 L (3)

¾v 2 f0; 1g; non-negative integer: ³MAX (4)


Proposed Routing StrategiesMinimize Maximum End-to-end Congestion (MEC)

Minimize 'MAX (5)

Subject to

X

v2Vs;d¾v = 1; 8s; d 2 N (6)

´v;v0 ¸ (¾v + ¾v

0 ¡ 1)£ qv;v0 ; 8v 2 V ; 8v0 2 V n Vs(v);d(v) (7)

ts;d +X

v2Vs;d

X

v02VnVs;d´v;v

0 £ ts(v0);d(v0) · 'MAX ; 8s; d 2 N (8)

¾v; ´v;v0 2 f0; 1g; non-negative integer: 'MAX (9)


Simulation ModelA conceptual view

networkspecifications

Kinput paths

optimizer

simulator

routing tablesprotocolbehavior

simulation results

1st. st

age

2nd. st

age

ILPformalization

in this study

K shortest paths with less links in common

K ∈ {2, 3, 4}

A set of K input pathsfor each pair of nodes


Simulation ModelThe OBS network global characteristics

Copenhagen

London Amsterdam Berlin

Brussels

Luxemburg

Prague

Paris

Zurich

Vienna

Milano

– W=16 λs / link

– 10 Gbit / channel

– Poisson pattern traffic

– Threshold-based bursts(100 KB)

– Symmetric traffic matrix

– JET signaling scheme

– Source routing decision(MCL / MEC)

– Full wavelength conversion

The COST 239topology


Simulation ModelThe adopted OBS architecture

Burst

CPHWDM Link

offset

EdgeRouter

CoreRouter

Optical Backbone

Access Layer


Simulation ModelNetwork topologies under study

ARPANET random12

NSFnet COST239


Simulation ModelSnapshot of COST239 *

* in OMNeT++


Simulation ModelSnapshot of ARPANET *

* in OMNeT++


Simulation ModelSnapshot of NSFnet *

* in OMNeT++


Simulation ModelSnapshots of submodules *

* in OMNeT++ actions in

wavelengthdimension

actions inspacedimension

edgeNode

coreNode


Simulation ModelStructure of the RoutingTable @ edgeNode

routingTable.h

class RoutingTable : public cSimpleModule{

private:struct RouteEntry {

std::vector<int> route;int numOfHops;bool accessed;

};

// container for Routing Tablestd::vector<RouteEntry> routingTable;

public:// ...

protected:// ...

}

abarra@ceotserver:~/cost239$ cat cost239_sp_s9.rte9 0 1 9 6 09 1 1 9 3 0 19 2 1 9 3 29 3 1 9 39 4 1 9 10 49 5 1 9 8 59 6 1 9 69 7 1 9 3 79 8 1 9 89 9 1 9 99 10 1 9 10abarra@ceotserver:~/cost239$

source routingis used


Simulation ModelBehavior of the coreChSched @ coreNode

handleMessage

// Function called for each event

if msg is a Control Packet HeaderprocessCph

else if msg is an Optical BurstprocessBurst

else, it is a selfmessageprocessHoldingTimer


Performance Evaluation

Simulationmodel

11 nodes26 b-links4.73 av-degreeW=1610Gbps/w11E+06 burstsL=100KBPoisson TPSymmetric TMFull Wconv



Simulationmodel

11 nodes26 b-links4.73 av-degreeW=1610Gbps/w11E+06 burstsL=100KBPoisson TPSymmetric TMFull Wconv


Conclusion● Both routing strategies (MCL and MEC) behave better than

the shortest path approach.

Ack

now

ledgm

ents

:

Thank you

Ending● An OMNeT++ simulation model was created for routing path

evaluation in OBS networks.● Two routing strategies for OBS networks are presented and

evaluated● It is possible to reduce burst loss taking preventive actions on

the space domain.● (Performance results may also depend on the network

connectivity..., further research is needed).● Network simulation model is being prepared for future work.


Optical Brust Switching network simulatin omnet

Documents

Transcript of Optical Brust Switching network simulatin omnet