Electrical and Computer Engineering University of Alberta 2nd Floor, 9107 – 116 Street

Factors Affecting the Efficiency of Demand-wise Shared Protection

Brian Forst, Wayne D. Grover

Contact: [email protected], [email protected]

Electrical and Computer EngineeringUniversity of Alberta2nd Floor, 9107 – 116 StreetEdmonton, Alberta, Canada T6G 2V4

Network Systems GroupTRLabs7th Floor, 9107 – 116 StreetEdmonton, Alberta, Canada T6G 2V4

2

October 9, 2007Factors Affecting the Efficiency of Demand-wise Shared Protection

DRCN 2007

OverviewOverview

• Background– 1+1 APS– DSP

• Methodology & Network Families• Design Results• Practical Methods to Increase Efficiency• Conclusion

3


DRCN 2007

OverviewOverview



4


DRCN 2007

1+1 Automatic Protection Switching (APS)1+1 Automatic Protection Switching (APS)

• Oldest and simplest protection architecture• Minimum 100% redundancy• Fastest protection

A BXWorking path

5


DRCN 2007

Why Demand-wise Shared Protection (DSP)?Why Demand-wise Shared Protection (DSP)?

• Renewed interest in 1:N DP APS as a network protection architecture

• Promising aspects include:– Fast failure-response times– Full pre-failure cross-connection of backup paths– No fault localization required– Conceptually simple to understand– Technically possible to operate in current networks– Possibility of low redundancy/high efficiency protection

• Motivation for this study:– In recent studies DSP is not exhibiting the low redundancy we

expected it could achieve

6


DRCN 2007

DSPDSP

A B

Working paths

1ab

ab

ds

k

7


DRCN 2007

DSPDSP

A B

X

1ab

ab

ds

k

8


DRCN 2007

DSPDSP

A BX

1ab

ab

ds

k

9


DRCN 2007

OverviewOverview



10


DRCN 2007

MethodologyMethodology

• Obtain DSP and APS design results– Hop-based costs– Distance-based costs

• Analyze route-cost statistics

• New ILP model and method to calculate route-cost statistics

11


DRCN 2007

Network TopologiesNetwork Topologies

• Two families of networks:

• Degree 3

• Degree 4

America Europe Germany

15 Node 20 Node 25 Node

12


DRCN 2007

OverviewOverview



13


DRCN 2007

Design ResultsDesign Results

• Hop-based costs– Degree 3 Networks: 1%, 1% and 7% savings relative to APS– Degree 4 Networks: 11% - 12% savings relative to APS

• Distance-based costs– Degree 3 Networks: 0.3%, 2.5% and 5.5% savings relative to APS– Degree 4 Networks: 8% - 9% savings relative to APS

• Here, hop-based costing gives greater reductions

14


DRCN 2007

OverviewOverview



15


DRCN 2007

What to do?What to do?

• Small savings over 1+1 APS found– Nowhere near SBPP values.

• Six areas of concern have been identified

16


DRCN 2007

1: Average Network Nodal Degree1: Average Network Nodal Degree

• Degree 4 networks have higher savings than degree 3– Greater amounts of disjoint routes

0%

10%

20%

30%

40%

50%

60%

70%

2 3 4 5 6Number of Disjoint Routes

Pro

port

ion

of D

isjo

int R

oute

s Degree 3 GroupDegree 4 Group

17


DRCN 2007

2: Demand Splitting Restrictions (Magnitude)2: Demand Splitting Restrictions (Magnitude)

• Integer amounts of demand– Cannot be further sub-divided

BAdAB = 2

BAdAB = 1

18


DRCN 2007

3: Demand Splitting Restrictions (Divisibility)3: Demand Splitting Restrictions (Divisibility)

APS DSP

dAB = 3 BA

w1 = 3

s = 3

Cost = 6

BA

w1 = 2

s = 2

w2 = 1

Cost = 5

dAB = 4 BA

w1 = 4

s = 4

Cost = 8

BA

w1 = 2

s = 2

w2 = 2

Cost = 6

19


DRCN 2007

4: Diminishing Returns of Greater Splitting4: Diminishing Returns of Greater Splitting

1ab

ab

ds

k

ab abs d2k

3k 2ab

ab

ds

50% reduction in spare capacity

4k 3ab

ab

ds

17% further reduction in spare capacity

20


DRCN 2007

5: Degree 2 Nodes and Chains5: Degree 2 Nodes and Chains

• Europe: 32% (9 of 28)• Germany: 41% (7 of 17)

• America: 15% (2 of 14)

A

1.5% Savings

6% SavingsGermany

21


DRCN 2007

5: Degree 2 Nodes and Chains5: Degree 2 Nodes and Chains

Europe Germany

22


DRCN 2007

6: Rapid Cost Increase of Higher Order Routes6: Rapid Cost Increase of Higher Order Routes

• Remember this equation?1

abab

ds

k

• We are fighting against it

• Two important ratios

23


DRCN 2007


• First ratio: Rn/R1

• Approximately 600 O-D pairs in each family

Degree 3 Families Degree 4 Families

100% R2/R1 = 1.55100% R2/R1 = 1.95

24


DRCN 2007


• First ratio: Rn/R1

• Approximately 600 O-D pairs in each family

Degree 3 Families Degree 4 Families

100% R2/R1 = 1.55100% R2/R1 = 1.95

100% R3/R1 = 2.31 50% R3/R1 = 3.55

40% R4/R1 = 3.23 1% R4/R1 = 3.55

6% R5/R1 = 4.85 0% R5/R1 = N/A

<1% R6/R1 = 9.10 0% R6/R1 = N/A

25


DRCN 2007


• Second ratio: Rn/(R1 + R2)

• R3 Case:

R1 : 2 work R2 : 2 spareCost = 2R1 + 2R2

R3 = 1 work

R1 = 1 work R2 = 1 spare

If R3 = R1 + R2

= R1 + R2 + (R1 + R2)

= 2R1 + 2R2

Cost

A BdAB = 2

26


DRCN 2007

0%

5%

10%

15%

20%

25%

30%

35%

40%

(0,0

.5]

(0.5

,0.6

67]

(0.6

67,0

.833

]

(0.8

33,1

]

(1,1

.5]

(1.5

,2]

(2,in

f)

Rn/ (R1+R2)

Freq of R3/(R1 + R2) RatioFreq of R4/(R1 + R2) Ratio


• Second ratio: Rn/(R1 + R2)

0%

5%

10%

15%

20%

25%

30%

35%

40%

(0,0

.5]

(0.5

,0.6

67]

(0.6

67,0

.833

]

(0.8

33,1

]

(1,1

.5]

(1.5

,2]

(2,in

f)

Rn/ (R1+R2)

Freq of R3/(R1 + R2) Ratio




0%

5%

10%

15%

20%

25%

30%

35%

40%

(0,0

.5]

(0.5

,0.6

67]

(0.6

67,0

.833

]

(0.8

33,1

]

(1,1

.5]

(1.5

,2]

(2,in

f)

Rn/ (R1+R2)





0%

5%

10%

15%

20%

25%

30%

35%

40%

(0,0

.5]

(0.5

,0.6

67]

(0.6

67,0

.833

]

(0.8

33,1

]

(1,1

.5]

(1.5

,2]

(2,in

f)

Rn/ (R1+R2)





Degree 3 Degree 4

27


DRCN 2007


• 6 disjoint routes

BA

25 Node

28


DRCN 2007


• Only 2 used in optimal solution!

AB

25 Node

First two routes: 400 kmThird route: 410 km

R3/(R1+R2) > 1

29


DRCN 2007

ConclusionConclusion

• New ILP model

• Identified important factors– Higher order route costs– Degree 2 nodes– Unit capacity (single-channel) demands

• Guideline for the application of DSP– 3-way splits

• These are new insights that give us an understanding of what limits the attainable efficiency of DSP

30


DRCN 2007

Questions?Questions?

Documents

Electrical and Computer Engineering University of Alberta 2nd Floor, 9107 – 116 Street