Академический Документы
Профессиональный Документы
Культура Документы
David Tipper
Associate Professor
Department of Information Science and
Telecommunications
University of Pittsburgh
Slides 6
http://www.sis.pitt.edu/~dtipper/2110.html
TELCOM 2110 5
1
Access Network Example
• Can make Access/Backbone analogy with transportation
networks
• Visit to Gauley River for whitewater rafting in WV
• The trip involves 4 segments:
– Travel from home in Pittsburgh via city streets to interstate
– Then traverse the 279, 79 Interstate backbone to the location
closest to Gauley River - U.S. Highway 19 exit
– Then you travel on U.S. highway 19 to get to Summersville, WV
(this is like a concentrator – is a four lane highway (with traffic
lights etc)
– Travel local state highways roads from Summersville to get to
Gauley River
TELCOM 2110 6
TELCOM 2110 7
2
Access Network Design
TELCOM 2110 9
3
Star Design
TELCOM 2110 10
One Concentrator
• N2 serves as a concentrator for N6 and N7.
• Shorter less expensive links are used from N6 and N7
• Cost=$8660; Maximum Utilization=46.4%
TELCOM 2110 11
4
Two Concentrators
• N2 for N6 and N7; N4 for N3.
• Cost=$8158; Maximum Utilization=46.4%
TELCOM 2110 12
MST Design
• Using MST algorithm, chooses N7, N4 as concentrators
• Cost=$7659; Maximum Utilization=46.4%
TELCOM 2110 13
5
MSTs Are Not Always Optimal Access Designs
TELCOM 2110 14
AUC EIR
• Can roughly categorize
HLR
access design problems
VLR
IBM
PCC ARD
SD
* 8x5
P 0
ETHER LIN K RS232C I NS ACT AL
M
RST
OOO13 0
AO N
6
PCC ARD
ALM
PWR AL
M FAN0 FAN1 PWR0 PWR1
BSC
BSC
telephone network
SD
* 8x5
P 0
ETHER LIN K RS232C I NS ACT AL
M
OOO13 0
RST
AO N
6
PCC ARD
ALM
PWR AL
M FAN0 FAN1 PWR0 PWR1
BSC
– Multi-center Design
BS7
BS5
BS7
BS5 BS6
BS6
• For example, cellular
networks with multiple base
station controllers
TELCOM 2110 15
6
One-speed One-Center Design
OR OR
TELCOM 2110 16
Approaches
7
One-speed One-Center Example
SPT(Star)
• Cost= $26358
• Very low link utilization and expensive
TELCOM 2110 19
8
MST
• Cost= $18,730
• More cost effective but has higher delays
TELCOM 2110 20
TELCOM 2110 21
9
Capacitated Minimum Spanning Tree (CMST)
• CMST problem:
Given
a center node N0
set of other nodes (N1, …, Nn),
set of weights(w1,…,wn) for each node,
the capacity of each link, W
cost matrix Cost(i,j),
• Find: a set of trees T1, …, Tk such that each Ni belongs to exactly
one Tj and each Tj contains N0 and the following holds
∑w
i∈T j ,i > 0
i < Wi
10
The Esau-Williams Algorithm
• Merger is allowed if the link capacity is not
exceeded – that is weight of nodes less than link
speed
weight(Comp(NK )) + weight(Comp(NJ )) ≤ W
TELCOM 2110 25
Esau-Williams Example
• W=3, each node has wi=1
• Tradeoff(1)=minj Cost(N1,NJ)-
Cost(Comp(N1),Center) Initial topology
=minj Cost(N1,N3) - (Comp(N1) dashed lines
contains N0)
=3-5= -2 (pick closest neighbor, N3) 2
• Tradeoff(2)=4-6= -2
• Tradeoff(3)=3-9= -6 8
5
6
• Tradeoff(4)=5-12= -7 4 12 4
• Tradeoff(5)=6-15= -9
7
12
6
• Tradeoff(5) is the smallest 0
15 5
9 6
• Accept link(5,3) merger to the solution
3
since weight constraint on component 5
10 8
tree with nodes 5 and 3 is not violated. 3
Σwi =w5+w3=2<=W=3
1
TELCOM 2110 26
11
Esau-Williams Example
• Next Iteration
– Tradeoff(1)=3-5= -2
– Tradeoff(2)=4-6= -2 Topology after 1
– Tradeoff(3)=3-9= -6 iteration
– Tradeoff(4)=5-12= -7 2
– Update Tradeoff(5)=7-9= -2
next shortest link out of 5 is (5,4) 8
5
6
(Comp(5)=9,node 5 goes through 4 12 4
node 3 to center)
7
– Tradeoff(5)=7-9= -2 12
6
15 5
0
• Pick Tradeoff(4) as smallest 9 6
• Accept (4,2) merger since 3
weight constraint on component 5
10 8
TELCOM 2110 27
Esau-Williams Example
• Next iteration
– Tradeoff(1)=3-5= -2 Topology after iteration 2
– Tradeoff(2)=4-6= -2
– Tradeoff(3)=3-9= -6 2
– Update Tradeoff(4)=6-6= 0
5
– Tradeoff(5)=7-9= -2 6
8
– Pick Tradeoff(3) 4 12 4
7
• Accept link (3,1) since 12
6
weight constraint on 0
15 5
component 9 6
not violated. 3
TELCOM 2110 28
12
Esau-Williams Example
• Next Iteration
– Tradeoff(1)=4-5= -1 Topology after iteration 3
– Tradeoff(2)=4-6= -2
– Tradeoff(3)=6-5= 1 which is Final topology
– Tradeoff(4)=6-6=0
– Since nodes 5 and 3 now go 2
through node 1 to Center,
update Tradeoff(5)=7-5=2 5
8
• Tradeoff(2) is lowest but 6
adding link(2,1) results a 4 12 4
component 7
with 4 nodes violate Σwi<=3. 12
6
• Reject(2,1) 0
15 5
recompute Tradeoff(2)=6-6=0
9 6
• Reject(1,2) similar reason.
Recompute Tradeoff(1)=5-5=0 10 8
3
5
• The access network is complete 3
TELCOM 2110 29
Example 2
0 1 2
1
3 4 5
1
6 7 8
TELCOM 2110 1 1 30
13
Example 2
Tradeoff(i)=minj Cost(Ni,NJ) -Cost(Comp(Ni),Center)
Tradeoff(1) = 1 -1 = 0
Tradeoff(2) = 1 -2 = -1
Tradeoff(3) = 1 – 1 = 0
Tradeoff(4) = 1 – sqrt(2) = -.414
Tradeoff(5) = 1 – sqrt(5) = -1.236
Tradeoff(6) =1-2 = -1
Tradeoff(7) = 1-sqrt(5) = -1.236
Tradeoff(8) = 1-sqrt(8) = - 1.828
Pick 8 to merge with either 7 or 5
0 1 2
Pick 7 since it has lower weight = 1 1
Checking capacity w7+w8 = 2 ≤ W = 3 4
3 5
1
6 7 8
1 1
TELCOM 2110 31
Example 2
Iteration 2
Tradeoff(1) = 1 -1 = 0
Tradeoff(2) = 1 -2 = -1
Tradeoff(3) = 1 – 1 = 0
Tradeoff(4) = 1 – sqrt(2) = -.414
Tradeoff(5) = 1 – sqrt(5) = -1.236
Tradeoff(6) =1-2 = -1
Tradeoff(7) = 1-sqrt(5) = -1.236
Tradeoff(8) = 1-sqrt(5) = - 1.236
Pick 5 to merge with node 2
0 1 2
Note node 4 or 8 merge is not allowed 1
by capacity constraint 4
3 5
Checking capacity w5+w2 = 3 ≤ W = 3 1
6 7 8
1 1
TELCOM 2110 32
14
Example 2
Iteration 3
Tradeoff(1) = 1 -1 = 0
Tradeoff(2) = 1 -2 = -1
Tradeoff(3) = 1 – 1 = 0
Tradeoff(4) = 1 – sqrt(2) = -.414
Tradeoff(5) = 1 – 2 = -1 – not allowed
Tradeoff(6) =1-2 = -1
Tradeoff(7) = 1-sqrt(5) = -1.236
Tradeoff(8) = 1-sqrt(5) = - 1.236
Pick 7 to merge with node 6
0 1 2
Note node 4 is not allowed by capacity 1
constraint 4
3 5
Checking capacity w6+w7 + w8= 3 ≤ W 1
6 7 8
1 1
TELCOM 2110 33
Example 2
Iteration 4
Tradeoff(1) = 1 -1 = 0
Tradeoff(2) = 1 -2 = -1 not allowed
Tradeoff(3) = 1 – 1 = 0
Tradeoff(4) = 1 – sqrt(2) = -.414
Tradeoff(5) = 1 – 2 = -1 not allowed
Tradeoff(6) =1-2 = -1 not allowed
Tradeoff(7) = 1-2 = -1 not allowed
Tradeoff(8) = 1-2 = -1 not allowed
Pick 4 to merge with node 3 or 1
0 1 2
Checking capacity w3+w4 = 3 ≤ W 1
Note all allowed merges have positive 4
3 5
Tradoffs so final topology with cost 10 1
6 7 8
1 1
TELCOM 2110 34
15
Esau-Williams Algorithm
TELCOM 2110 35
• Algorithm does
as well if the sites
have a variety of
different traffic.
• Links are
9600bps
• 50% of sites
require 2400bps
• Others require
4800bps
TELCOM 2110 36
16
Line Crossings in Access Designs
TELCOM 2110 37
Sharma’s Algorithm
TELCOM 2110 38
17
Example of Sharma’s algorithm
0 1 2
1
3 4 5
1
6 7 8
TELCOM 2110 1 1 39
1 1
TELCOM 2110 40
18
Example of Sharma’s algorithm
1 1
TELCOM 2110 41
19
Sharma’s Algorithm Design
• Cost= $16021, Sfirst = N17
TELCOM 2110 43
TELCOM 2110 44
20
Sharma vs. Esau-Williams
• EW_Ratio=SharmaCost/EWCost;
S_Ratio=EWCost/SharmaCost
• In general use Esau- Williams unless require no lines cross
TELCOM 2110 45
Access Design
VLR
speed one center
MSC design
Bay Networks
* 8x5
P
AO N
6
0
OOO13 0
RST
PW R
ETHER
AL
M
LIN K RS232C
ALM
FAN0 FAN1 PWR0 PW R1
I NS ACT AL
M
PCC ARD
Centillion 1400
SD
Bay Networks
* 8x5
P
6
0
OOO13
AO N
0
RST
PWR
ETHER
AL
M
LIN K RS232C
ALM
FAN0 FAN1 PWR0 PWR1
I NS ACT AL
M
PCC ARD
Centillion 1400
SD
• Local loop in
telephone network
BSC
BSC
SD
* 8x5
P 0
ETHER LIN K RS232C I NS ACT AL
M
OOO13 0
RST
AO N
6
PCC ARD
ALM
PWR AL
M FAN0 FAN1 PWR0 PWR1
• Cellular network
BSC
BS3 BS2
BS3
BS4
connecting BS to BSC
BS2
BS1
BS4 BS1
• LAN, host to hub or
switch
BS7
BS5
BS7
BS5 BS6
BS6
• Two algorithms
1. Esau-Williams
2. Sharma
TELCOM 2110 46
21