1 Distributed Computing Optical networks: switching cost and traffic grooming Shmuel Zaks...

Post on 24-Dec-2015

217 views 0 download

Tags:

Transcript of 1 Distributed Computing Optical networks: switching cost and traffic grooming Shmuel Zaks...

1

Distributed Computing Optical networks: switching cost and traffic grooming

Shmuel Zaks

zaks@cs.technion.ac.il

©

2

OutlineOutline

Optical networksOptical networks ModelModel The Min ADM ProblemThe Min ADM Problem The Traffic Grooming The Traffic Grooming

ProblemProblem Algorithm GROOMBYSCAlgorithm GROOMBYSC

3

the fiber serves as a transmission medium

Electronic switch

Optic fiber

Optical networks - 1st generation

4

Routing in the optical domainTwo complementing technologies:- Wavelength Division Multiplexing (WDM):

Transmission of data simultaneously at multiple wavelengths over same fiber- Optical switches: the output port is determined according to the input port and the wavelength

Optical networks - 2nd generation

5

Wavelength Division Multiplexing (WDM)

Directed:1234

Symmetric:1234

Undirected:1234

Optic Fiber

Optic Fiber

Optic Fiber

6

Optical Switches

No two inputs with the same wavelength should be routed on the same edge.

7

Lightpaths

ADM

ADM

Data in electronic form

Data in electronic form

8

A virtual topology

9

Lightpaths

p1

p2

1 2( ) ( )w p w p

Valid coloring

10

The Routing Problem Input :

A graph G=(V,E) A set or sequence of node pairs (ai,bi)

Output: A set or sequence of paths pi =(ai, v1, …,

bi)

11

The Load Given a graph G=(V,E) and a set P of

paths on the graph, we define: for any edge e of the graph:

the load on this edge l(e)=|Pe|

The (maximum, minimum, average) load on the network:

| eP p P e P

max

min

max ( ) |

min ( ) |

( ) avge E

L L l e e E

L l e e E

L l e E

12

Wavelength Assignment Problem (WLA) Input:

A graph G=(V,E). A set or sequence of paths P.

Output: A coloring w of the paths:

Constraint::w P Na

, , ' , ' ( ) ( ')ee E p p P p p w p w p

13

Routing and WLA (RLA/WRA)

Input : A graph G=(V,E) A set or sequence of node pairs (ai,bi)

Output: A set or sequence of paths pi =(ai, v1, …, bi) A coloring w of the paths: Constraint:

:w P N, , ' , ' ( ) ( ')ee E p p P p p w p w p

14

Cost Measure: # of colors

For any legal coloring w of the paths:

Range( ) ( ) ( ) |W w w P w p p P

15

Optimization Problems Goal:

MINW: Minimize W. or MAXPC: Maximize |Domain(w) | under

the constraint W<=Wmax.

16

Static vs. Dynamic vs. Incremental

Static: The input is a set (of pairs or paths), the algorithm calculates its output based on the input.

Incremental (Online): The input is a sequence of input elements (pairs or

paths). It is supplied to the algorithm one element at a time. The output corresponding to the input element is

calculated w/o knowledge of the subsequent input elements.

Dynamic: Similar to incremental The sequence may contain deletion requests for

previous elements.

17

WLA (A trivial lower bound) For any instance of the WLA

problem: W>=L.

Proof: Consider an edge e, such that L=l(e). There are L paths p1, …, p|L| using e,

because the paths are simple. Therefore :

, | |, ( ) ( )i ji j L w p w p

( ) |1iw p i L L

18

WLA (A trivial lower bound) For some instances W > L.

L=2

W=3

19

Static WLA in Line Graphs The GREEDY algorithm: // The set of integers for i = 1 to |V| do

for each path p=(x,i) do for each path p=(i,x) do

( )W W w p

{ ( ) min ;

\ ( )}

w p W

W W w p

W N

20

Static WLA in Line Graphs Correctness, obvious. Optimality: By induction, After node i is processed, the

claim is correct, i.e. Where

W(i) is the value of after node i is pocessed, and

L(i) is the maximum load on the edges processed so far.

( ) ( )W i L i

max WN

21

OutlineOutline

Optical networksOptical networks Model Model The Min ADM ProblemThe Min ADM Problem The Traffic Grooming The Traffic Grooming

ProblemProblem Algorithm GROOMBYSCAlgorithm GROOMBYSC

22

Electronic ADM

23

number of wavelengthsSwitching cost

ADM

24

The MIN ADM Problem

W=2, ADM=4 W=1, ADM=3

25

The Goal

Given a set of lightpaths, find a valid coloring with minimum number of ADMs.

26

Static WLA in Line Graphs Note: After a slight modification, the Greedy

algorithm solves optimally the MINADM problem too:

At each node, first use the colors added to at this step.

It’s straigtforward to show that this: Does not harm the optimality w.r. to the MINW prb. Solves the MINADM problem optimally at each node.

W

27

Static WLA in Line Graphs The GREEDY algorithm: // The set of integers for i = 1 to |V| do

for each path p=(x,i) do for each path p=(i,x) do

( )W W w p

{ ( ) min ;

\ ( )}

w p W

W W w p

W N

28

W-ADM tradeoff

W=2, ADM=8 W=3, ADM=7

29

ring (Eilam, Moran, Zaks, 2002) reduction from coloring of circular arc

graphs.

NP-complete

Minimizing # of ADMs –

Gerstel, Lin, Sasaki, 1998

cs

30

Coloring of Circular arc Graphs

Consider: a ring H (the host graph) and A set of paths P in H.

The graph G=(P,E) constructed as follows is a circular arc graph:

There is an edge (p1,p2) in e if and only if p1 and p2 have a common edge in H.

The problem of finding the chromatic number of a circular arc graph is NP-Hard [Tuc 75’]

31

The reduction

The min W problem is exactly the circular arc coloring problem. But we will show NP-hardness even of the special case L=Lmin.

Given an instance C,P where C is the ring and P is the set of paths, we construct an instance C, P’ (by adding paths to P) such that Lmin(P’)=L(P’)=L(P).

32

The reduction (cont’d) Claim: P is L-colorable iff P’ is L-

colorable.

Claim: ADM(P)=ADM(P’).

Therefore w.l.o.g. all the edges have the same load (L).

33

|ADMs|=7=7+0

|ADMs|=9=6+3

|ADMs| = N + |chains|

Basic observationN lightpaths

cycles

chains

34

The reduction (cont’d)

P’ is L-colorable iff

P’ can be partitioned into L cycles iff

ADM(P’)=|P’|.

35

R ALG 2R R OPT 2R

ALG 2 x OPT

R: # of lightpaths ALG: # of ADMs used by the algorithm OPT: # of ADMs used by optimal solution

Approximation algorithms

36

3/2 - Calinescu, Wan , 2002 10/7+ - Shalom, Z. , 2004 10/7 - Epstein, Levin, 2004

ALG 2 x OPT

Approximation algorithms

37

OutlineOutline

Optical networksOptical networks Model Model The Min ADM ProblemThe Min ADM Problem The Traffic Grooming The Traffic Grooming

ProblemProblem Algorithm GROOMBYSCAlgorithm GROOMBYSC

38

The Traffic Grooming Problem

A generalization of the MIN ADM problem.

Instead of requests for entire lightpaths, the input contains requests for integer multiples of 1/g of one lighpath’s bandwidth.

g is an integer given with the instance.

39

The Traffic Grooming Problem

W=2, ADM=8 W=1, ADM=7

g=2

40

The Goal

Given a set of requests and a grooming factor g, find a valid coloring with minimum number of ADMs.

41

Notation & Immediate Results R: The # of requests. SOL: The # of ADMs used by a

solution. OPT: The # of ADMs used by an

optimal solution.R/g SOL 2RR/g OPT 2RSOL = SOL/OPT 2g

42

OutlineOutline

Optical networksOptical networks ModelModel The Min ADM ProblemThe Min ADM Problem The Traffic Grooming The Traffic Grooming

Problem Problem Algorithm GROOMBYSCAlgorithm GROOMBYSC

43

Main Resultg > 1, Ring Networks:

General traffic:

An O(log g) approximation algorithm for any fixed g.

Can be used in general networks

Analysis can be extended to some other topologies.

44

Approximation algorithm (log g)

¬ ÈS S {A}

Input: Graph G, set of lightpaths P, g > 0

Step 1: Choose a parameter k = k(g).

Step 2: Consider all subsets of P of size

If a subset A is 1-colorable (i.e., any edge is used at most g times) then

weight[A]=endpoints(A);

£ ×k g

45

Algorithm (cont’d)

Step 3: COVER(an approximation to) the Minimum Weight Set Cover of S[], weight[], using [Chvatal79]

Step 4: Convert COVER to a PARTITION

PARTITION induces a coloring of the paths

46

Analysis

Let , then:

If B is 1-colorable then A is 1-colorable (correctness).

Cost(A) Cost(B).

A B

Therefore: …

47

k g

cost(PARTI TI ON)

weight(COVER)

H weight(MI NCOVER)

(1+ln(k g))w

ALG=

Sh Ceig t( )

for every set cover SC.

48

Lemma: There is a set cover SC, s.t.: 2g

weight( ) 1+SC Pk

O T

(1+ln(k g)) weightALG (SC)

for any set cover SC.

49

k g

weight(COVER)

H weight(MI NCOVER)

(1+ln(k g))weight( )

2g(1+ln(k g)

A

) 1+

SC

k

LG

OPT

Conclusion:

For k = g ln g : 2lng+o(lngA G )L OPT

50

Proof of Lemma

Lemma: There is a set cover SC, s.t.: 2g

weight( ) 1+SC Pk

O T

51

Proof of LemmaConsider a color of OPT.Consider the set P of paths colored .Consider the set of ADMs operating at wavelength . (i.e. endpoints(P) )Divide endpoints(P) into sets of k consecutive nodes.For simplicity assume |endpoints(P)|=m.k

52

k k k k

1weight[S ] k+gS1 S2 Sm

M=4 k=6

53

Analysis (cont’d),

,1

,1

[ ]

[ ] ( )

.

[ ] 1

i

m

ii

m

ii

weight S k g

weight S m k g

OPT m k

gweight S OPT

k

w/o the assumption we have:

,1

2[ ] 1

m

ii

gweight S OPT

k

,1

2[ ] 1

m

ii

gweight S OPT

k

54

Analysis (cont’d)

,iS S

,iS S

, ,, [ ] endpoints( ) .i ii weight S S k g

thus

,,

ii

P S

Moreover

,iSC S Therefore

Is a set cover considered by the algorithm.