Routing and Wavelength Assignment in Wavelength-

Convertible Waveband-Switched Networks

Routing and Wavelength Assignment in Wavelength-

Convertible Waveband-Switched Networks

Fang-Sheng Lin*

Ching-Fang Hsu

Te-Lung Liu

2

Outline• Introduction• Related Work

– Reconfigurable MG-OXC architecture– Waveband Assignment with Path-Graph

(WAPG) algorithm• The Proposed Scheme

– Problem Definition– Least-Configuration with Bounded Conversion

(LCBC) algorithm• Performance Evaluation• Conclusions

3

Introduction (1/5)

• Owning to the development of DWDM systems, the number of wavelengths gets larger and larger.

• Traditional OXCs which switch traffic only at wavelength granularity will need a great deal of wavelength ports.– higher complexity – difficulty associated with controlling and

management of such large OXCs

4

Introduction (2/5)

• The main idea of Waveband Switching (WBS) is to aggregate a set of wavelengths into a band and switch the band using a single port whenever possible.

5

Introduction (3/5)

• The wavelength continuity constraint on lightpath establishment existing in wavelength routed networks also apply to WBS networks.

• One possible way to relax the wavelength continuity constraint is to use wavelength converters at the switching node.

6

Introduction (4/5)

• It is more practical and cost-effective to share a set of limited range wavelength converters at each node.

7

Introduction (5/5)

• Focused on sub-path grouping strategy, we proposed a new heuristic algorithm to solve the dynamic RWA problem of wavelength-convertible WBS networks efficiently.

8

Related Work (1/4)- Reconfigurable MG-OXC

• The Multi-granular Optical Cross-Connect (MG-OXC) architecture with wavelength conversion bank [1]

FXC

BXC

WXC

…

…

…

…

…

……

…

…

…

…

…

… …

X

βY

αX

BTF

BTW WTB

FTB

WXCLayer

FXCLayer

BXCLayer

Y

Wavelength Conversion Bank

9

Related Work (2/4)- Reconfigurable MG-OXC

• There is a unique issue related to using wavelength converters:– In WBS networks, an instantiation of

wavelength conversion requires all wavelengths in a waveband to be de-multiplexed and results in extra ports consumption.

• Hence, inefficient banding and employment of wavelength converters may cause more blocking of future requests due to the limitation of the OXC ports.

10

Related Work (3/4)- WAPG algorithm

• The authors in [1] proposed a heuristic algorithm called Waveband Assignment with Path-Graph (WAPG) to address the effect on the blocking performance and efficient usage of wavelength converters in WBS networks.

11

Related Work (4/4)- WAPG algorithm

• WAPG– Find out a wavelength-continuous path

similar First-Fit algorithm first.– If no wavelength-continuous path can

be found, find a non-wavelength-continuous path using minimum number of converters.

• However, it may cause undesired port usage.

12

The Proposed Scheme (1/5)- Problem definition

• The network topology– G=(V, E)

• Each fiber link has W wavelengths, which are partitioned into B uniform wavebands.

• According to the index continuity, a fixed number K of wavelengths are chosen to be grouped into a band

13

The Proposed Scheme (2/5)- Problem definition

• For the wavelength-convertible WBS networks, the major objectives of dynamic RWA problem include– to minimize the configuration (switching) cost,

i.e., the total number of ports used,– to utilize wavelength conversion in a most

efficient manner, and– to achieve a magnificent network performance

at the same time.

14

The Proposed Scheme (3/5)- Least-Configuration with Bounded Conversion

• We proposed a new heuristic algorithm, called Least-Configuration with Bounded Conversion (LCBC), based on– the fixed routing algorithm,– the layered graph approach, and– a well-designed cost function

15

The Proposed Scheme (4/5)- Least-Configuration with Bounded Conversion

s d

0v 1v 2v 3v

Layer 0

Layer 1

Layer 2

Layer W-1

….

10

Wv 11

Wv 13

Wv12

Wv

00v

01v

03v

02v

10v

11v

13v

12v

20v

21v

23v

22v

An illustration of layered graph modeling

's 'd

transmission edge

conversion edge

16

• The proposed cost function

The Proposed Scheme (5/5)- Least-Configuration with Bounded Conversion

W

DWD vv 1,,,, ',',

0, no wavelength conversion performed

1, wavelength conversion occurredψ=

The extra port consumption at node v while λ is the input channel and λ' is the output channel [9]

Wavelength

conversion degree

Link capacity

The weight factor

17

Performance Evaluation (1/4)

• USAnet – 46 nodes and 76 links.

• Every node is assumed to be a MG-OXC. • Adopting the share-per-node architecture

– 25 wavelengths converters per node

18

Performance Evaluation (2/4)

• Traffic pattern– A Poisson process with mean γ– Exponentially distributed connection

duration time whose mean is 1 time unit – All existing connections can not be

rearranged.

19

Performance Evaluation (3/4)

β vs. blocking probability (W=80, K=10)

93.8%

3.8%

20

Performance Evaluation (4/4)

Blocking probability vs. arrival rate(W=80, K=10, β=0.75)

51.1%

26%

97.8%

21

Conclusions (1/2)• In this paper, we proposed a heuristic algorithm,

named LCBC, to solve the problem of dynamic RWA in wavelength-convertible WBS networks.

• Adopting fixed routing as the routing selection algorithm, we transform the wavelength selection problem into an equivalent shortest-path problem in an auxiliary graph.

• Moreover, we proposed a cost function to calculate an appropriate weight to each edge, such that limited resource, including BTW/WTB ports and wavelength converters, could be utilized more efficiently.

22

Conclusions (2/2)

• To investigate the efficiency, we developed the simulation to observe the performance of LCBC and that of WAPG with various β, and conversion degree.

• LCBC can achieve much significant blocking performance gain.

23