Documente Academic
Documente Profesional
Documente Cultură
1
WAN Network Design
•Design Variables
• Network Topology (possibly facility location as well)
• Channel Capacity
• Routing Policy
•Performance Metrics
depends on network application and layer
• Circuit Switched Network
•Call Blocking, Availability
• Packet network
• Delay
• Delay Jitter
• Throughput
• Packet Loss
2
Example of Tradeoffs/Designs
Example Design 1
Tree type of design, cost reduced to $133,584/month.
The average number of hops, 7.84, is high.
Design has only high speed links (T1 and 256Kbps lines)
Poor reliability
3
Example Design 2
Two level design – backbone with edge nodes
Data centers and servers are interior nodes of the backbone tree.
Cost reduced to $96,777; average hops= 3.41
Reliability poor – if backbone link failure – large impact
Example Design 3
Instead of tree, interior (backbone) nodes are connected with high-speed
links to form a 2-connected graph.
Cost = $ 112,587/month, better reliability
4
Example Design 4
Alternate 2 connected backbone design with
slightly lower cost $112,587Æ$108,724 per month
5
Example Design 5
Example Design 6
Expanding the backbone further by additing concentrators at
N4, N10, and N13; slightly reduces the cost =
$101,806/month
6
Examples of Real WAN Backbone Networks
Worldcom: http://www1.worldcom.com/global/about/network/maps/
MCI/UUNET
Sprint IP Backbone
7
Example of Real WAN Networks
British
Telecom
Backbone
8
Network Layers
Traffic at higher layers is demand at lower layers
DCS
Circuit OCn
Switched
Optical layer Transport
Layer
9
Mesh Network Design Algorithms
10
Mentor Algorithm Step 1
• Choose backbone sites. (Threshold Cluster Algorithm)
• Calculate the normalized weight NW(Ni)=W(Ni)/C
• In NW(Ni) > WPARM, select Ni as a backbone node,
• Group end sites Nj around a backbone site, Ni, based on
Cost(Nj, Ni)/MAXCOST < RPARM.
• Where MAXCOST=Max i,j Cost(Nj, Ni)
Merit function – gives equal value to proximity to center of network and weight
merit(n)= 0.5(MaxDistCtr-distCtrn)/MaxDistCtr + 0.5(Wn/W_Max)
Here
DistCtrn = ( xn − xctr ) 2 + ( yn − yctr ) 2
∑
x × Weight
n n
xctr = n∈ N
and MaxDistCtr = max ( xn − xctr ) 2 + ( yn − yctr ) 2 ∑ Weight
n∈N n
n∈ N
Center of Mass (xctr, yctr) defined by ∑ y × Weight
n n
yctr = n∈ N
largest merit get picked as backbone node. Group end node around it. Repeat until
all nodes are covered in groups.
11
Mentor Algorithm Step 2
12
Mentor Algorithm Step 4
•Sequencing node pairs to prepare adding additional direct links to the tree.
Use the tree to list node pair in “sequence”.
The node pair with longer path will be listed first – OUTSIDE IN ORDERING
For Example
13
Mentor Algorithm Step 6
• Decide which node pairs deserve direct links.
• Start with the top node pair (N1,N2) in the sequence.
• Calculate the number of basic links needed
n=ceil(Traf(N1,N2)/C).
• Compute resulting link utlization u=Traf(N1,N2)/(n*C)
• If u > utilmin, add direct link between N1 and N2.
• If u < utilmin, do not add direct link, but instead direct
traffic 1 hop through the tree,
• Add Traf(N1,N2) to Traf(N1,H) and Traf(H,N2). Here H is the
home of (N1,N2).
• Remove (N1,N2) from the sequence and repeat Step 6 again until all
node pairs are processed.
• Idea is to aggregate traffic to justify links connecting sites several hops
apart in tree
TELCOM 2110 Spring 06 28
9 The smaller the value of utilmin, the easier it is to add direct links.
• We then pick the best design from the set
• Note previous example designs 1-6 made with mentor
algorithm with different parameter settings
TELCOM 2110 Spring 06 29
14
Example of Mentor Algorithm
15
Example of Mentor Algorithm
Same 5 backbone nodes but with α=0.1, utilmin=0.9
Cost = $209,220/month.
16
Cost vs. Size of Backbone
Reliability/Survivability
17
2-Connected Backbones
MENTour
18
Example
45 Node Example used earlier – Example 1-6
The initial topology for MENTour Example 2 Design
Example
A final design by MENTour has lower cost
19
WAN Packet Network Design
20
Delay Calculations
Delay Calculations
γ = ∑ γ sd
s ,d
M M
Fi 1 Fi
T =∑ Di = ∑ (C − F )
i =1 γ γ i =1 i i
21
Delay Calculations
Ds ,d = ∑D
i∈ path ( s , d )
i
f (T ) or ∑ cost(C )
i
i
22
Solution to Minimization Problem
Yields
∂Q
=0 …M equations
∂C i
∂Q
=0 1 equation from constraint
∂α
23
Solution to Minimization Problem
Yields M
Fi
C i = Fi + (C − ∑ Fi ) M
i =1
∑
i =1
Fi
for i = 1, 2,…, M
Fi is the minimum Ci since ρi = Fi/Ci < 1
M
24
Example
Consider mesh topology below
links are numbered for reference
traffic matrix gives mean rate from source to destination in packets per
second,
average packet length = 1000 bits
budget for link bandwidth results in C = 48 Kbps
Routing/Flows
25
Routing/Flows
Link PPS Fi Ci Di
Routing/Flows
Link PPS Fi Ci Di
26
Min-Max Capacity
Link Ci Di
1 6418.6 .306
2 6418.6 .306
3 3398.6 .306
4 6908.6 .306
5 4088.6 .306
6 7148.6 .306
7 13218.6 .306
27
Cost Constraint Functions
More realistic Cost constraint.
• Fixed cost per link, ai.
• Distance-based or other variable factor per link, di
M
J = ∑ ( d i C i + ai )
i =1
• Available cost Ja
M
J a = J − ∑ ( d i Fi + a i )
i =1
Solution
∑i =1
di
28
Example
Example
29
Minimum Delay Example
Min-Max Example
J
C i = Fi + M
a
∑
i =1
di
30
WAN Packet Design
Heuristic Algorithm
Heuristic
Begin
Let D = set of traffic demands or flows;
Let E = set of all potential links;
do {
randomly select an order of traffic demands in D to be routed;
for each traffic demand k in the order
{
if (traffic demand k has an existing route)
temporarily remove its required bandwidth along the route;
End
TELCOM 2110 Spring 06 63
31
Summary
32