1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path...
-
Upload
emmeline-fox -
Category
Documents
-
view
221 -
download
1
Transcript of 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path...
![Page 1: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/1.jpg)
1
EL736 Communications Networks II: Design and Algorithms
Class8: Networks with Shortest-Path Routing Yong Liu
10/31/2007
![Page 2: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/2.jpg)
2
Outline
Shortest-path routing MIP Formulation Duality and Shortest-Path Routing Heuristic Method for link weights Examples Extensions
![Page 3: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/3.jpg)
3
Shortest-path Routing Take the shortest-path(s) from one point to
the other path length = summation of link weights algorithm: Dijkstra, Bellman-Ford, extensions, intra-domain routing: link state: OSPF, IS-IS equal-cost multi-path split (ECMP)
Intra-domain Traffic Engineering Good end-to-end performance for users Efficient use of the network resources Reliable system even in the presence of failures
![Page 4: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/4.jpg)
4
TE Optimization: The Problem
Intra-domain Traffic Engineering Predict influence of weight changes on
traffic flow Minimize objective function (say, of link
utilization) Inputs
Networks topology: capacitated, directed graph
Routing configuration: routing weight for each link
Traffic matrix: offered load each pair of nodes
Outputs Shortest path(s) for each node pair Volume of traffic on each link in the graph Value of the objective function
1
2
3
4
5
6
7
![Page 5: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/5.jpg)
5
Which link weight system to use Link Weight can be
1, (hop count) propagation delay (const.) 1/C (Cisco) congestion delay (load sensitive, online
update)
Objective dependent choice hop count vs. congestion delay ECMP vs. equal delay routing
1
2
3
4
100
100 100
100
801
2
3
4
100
100 100
100
80 100
![Page 6: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/6.jpg)
6
Shortest Path Routing: bounded link delay
![Page 7: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/7.jpg)
7
Penalty Function
use link penalty function to replace link constraints
![Page 8: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/8.jpg)
8
Shortest Path Routing: minimum average delay
load sensitive link delay
piece-wise linear approximation
![Page 9: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/9.jpg)
9
Shortest Path Routing: minimum average delay
![Page 10: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/10.jpg)
10
Minimization of Maximum Link Utilization
![Page 11: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/11.jpg)
11
MIP Formulation
![Page 12: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/12.jpg)
12
MIP Formulation
![Page 13: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/13.jpg)
13
Duality: Lagrangian
Slides from Convex Optimization, Boyd & Vandenberghe
![Page 14: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/14.jpg)
14
Duality: dual function
Slides from Convex Optimization, Boyd & Vandenberghe
![Page 15: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/15.jpg)
15
Dual Problem
Slides from Convex Optimization, Boyd & Vandenberghe
![Page 16: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/16.jpg)
16
Duality Theorem Weak Duality:
always hold (convex, non-convex problems) find non-trivial lower bounds for complex problems duality gap:
Strong Duality: does not hold in general hold for most convex problems, (including LP) zero duality gap, obtain optimal solution for the original
problem by solving the dual problem. Advantages of working with Duals
less constraints decoupling distributed algorithms:
• distributed routing algorithms• end system congestion control, TCP
![Page 17: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/17.jpg)
17
Duality: routing example
A Bh
f1(x1), x1
f2(x2), x2,
, h-x1-x2
Lagrange dual function
Decoupling
minimal delay routing
![Page 18: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/18.jpg)
18
Duality: routing example
A Bh
f1(x1*), x1
*
f2(x2*), x2
*,
*, h-x1*-x2
*=0
Dual algorithm: increase delay on virtual link if x1+x2<h, decrease
delay otherwise
Dual Problem
Strong Duality
![Page 19: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/19.jpg)
19
Routing Duality: generalization
multi-demand/multi-path
routing duality
optimal flows only on shortest-paths!
![Page 20: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/20.jpg)
20
Duality and Shortest-path Routing
![Page 21: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/21.jpg)
21
Dual Formulation
Duality
optimal flows only on shortest-paths!
![Page 22: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/22.jpg)
22
Optimal Link Weights
use optimal multipliers as link weights non-zero flows only on shortest paths ECMP Optimal Flow Allocation good solution if most demand pairs only
have one shortest path.
![Page 23: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/23.jpg)
23
Heuristic Methods Weight Adjustment
iterative local search increase weights for over-loaded links, decrease weights for under-
loaded links adjust weights for more balanced allocation
Simulated Annealing random initial link weights explore neighborhood: pick a random link, increase/decrease its
weight by one annealing: move to a worse weight setting with decreasing
probability Lagrangian Relaxation (LR)-Based Dual Approach
optimum Lagrange multipliers lead to optimal solution iterative algorithm to solve problem in dual space Step1: given a set of multipliers, obtain link weights, and shortest
path flow allocation Step2: adjust multipliers according to link rates and link capacities,
go back to step1 if stopping criteria not satisfied.
![Page 24: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/24.jpg)
24
Example: impact of different link weight systems
AT&T 90-node WorldNet IP Backbone scaled up demand volumes
average delay maximum link utilization
![Page 25: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/25.jpg)
25
Extensions Un-capacitated Shortest-Path Routing
modularized dimensioning
Optimizing link weights under transient failures 50% network failures < 1 min; 80% < 10 min.s no time to re-compute weights after each failure good weight setting for both normal and failure
situation
Selfish Routing and Optimal Routing every user choose minimum delay path Nash Equilibrium vs. Social Optimum
A B
f1(x1), x1
f2(x2), x2,
f3(x3), x3,
h
![Page 26: 1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.](https://reader038.fdocuments.us/reader038/viewer/2022103122/56649cf95503460f949c9f01/html5/thumbnails/26.jpg)
26
Braess Paradox
adding a link increase user delay
1
2
3
4
x
1 x
1
11
2
3
4
x
1 x
1
1
delay=1.5 delay=2
0