Download presentation
Presentation is loading. Please wait.
Published byJob Blair Modified over 6 years ago
1
Multipath Routing for Congestion Minimization & Multiroute Flows
Chandra Chekuri Univ of Illinois, Urbana-Champaign Joint work with Mark Idleman
2
s-t flows in directed graphs
G=(V,E) directed graph with edge capacities c(e), e in E Edge based definition of flow f(e) flow on edge e flow conserved at internal nodes V \ {s,t} f(e) β€ c(e) for all e Path based definition of flow f(p) flow on path p β π«(st) πβπ π π β€π π for all e
3
Flow Decomposition Edge flow f of value v can be decomposed into a path flow in O(nm) time A path flow of value v induces an edge flow of same value
4
Multi-route flows π«(st) = { p | p is a st path } π«(st, h) = {p = (p1,p2,...,ph) | each pj β π«(st) and the paths are edge-disjoint } h-route s-t flow f : π«(st, h) β R+ f(p) flow on path-tuple p
5
s p q t
6
Multiroute flows: basic theorem
[Kishimoto,Aggarwal-Orlin] Theorem: An acyclic edge s-t flow x : E β R+ with value v can be decomposed into a h-route flow iff x(e) β€ v/h for all edges e 3 2 1 1 s t s t 1 1
7
Congestion Minimization
Choose a path for each pair Minimize max number of paths using any edge (congestion) G
8
Congestion Minimization
0.7 Choose a path for each pair Minimize max number of paths using any edge (congestion) Special case: Edge-Disjoint Paths [Raghavan-Thompsonβ87] Solve mc-flow relaxation (LP) Randomly pick a path according to fractional solution Chernoff bounds to show approx ratio of O(log n/log log n) 0.5 0.3 0.25 G 0.1 0.15 0.2 0.15 0.65
9
Multipath Routing h2 = 1 s3 s2 s1 t1 t2 t3
Choose hi paths for pair (si, ti) (assume paths for pair disjoint) Minimize max number of paths using any edge (congestion) G h1 = 2 h3 = 2
10
Multipath Routing s3 s2 s1 t1 t2 t3 0.7
Choose hi paths for pair (si, ti) (assume paths for pair disjoint) Minimize max number of paths using any edge (congestion) [Srinivasanβ99] Solve relaxation (LP) Dependent rounding O(log n/log log n) approx via negative correlation 0.5 0.3 0.3 G 0.25 0.95 0.7 0.5 0.8
11
Can use multiroute flow decomposition
Decompose for each pair flow into hi-route flow of value 1 Do independent rounding ala Raghavan-Thompson 0.5 0.3 s1 t1 0.25 0.95 0.2 0.25 0.05 0.5
12
Advantages Rounding and analysis exactly the same as it is for standard congestion minimization. No need for technicalities of dependent rounding. If paths are short can use Lovasz-Local-Lemma to reduce congestion to O(log d/log log d) where d is max # of edges in any path collection of decomposition (via existing results of [Srinivasan, Moser-Tardos, Haeupler-Saha-Srinivasan])
13
Choosing Paths G=(V,E) , (s1,t1), β¦, (sk,tk) and need hi edge-disjoint paths per pair Minimize congestion
14
Multi-route flow LP Min C ο₯ p β P(siti, h) f(p) β₯ 1 for all (si, ti)
ο₯ p β P(siti, h) :e β p f(p) β€ C for all e, i f(p) β₯ 0 for all p
15
Multi-route flow LP Min C ο₯ p β P(siti, hi) f(p) β₯ 1 for all (si, ti)
ο₯ p β P(siti, hi) :e β p f(p) β€ C for all e, i f(p) β₯ 0 for all p How do we solve LP? Ellipsoid for dual: separation oracle is min-cost flow Compact LP followed by multi-route flow decomposition
16
Compact LP Min C for each i flow of value hi from si to ti f(e,i) β€ 1 for all e, i βi f(e,i) β€ C for all e f(e,i) β₯ 0 for all e, i Can decompose f(e,i) into multi-route hi-route-flow and choose one βpath collectionβ for each (si,ti) independently
17
Compact LP Min C for each i flow of value hi from si to ti
f(e,i) β€ 1 for all e, i βi f(e,i) β€ C for all e f(e,i) β₯ 0 for all e, i [Doerr-Kunnemann-Wahlstromβ10] solve same LP but do βstandardβ flow decomposition which does not guarantee disjoint paths. They employ dependent rounding after standard flow decomposition.
18
Choosing Short Paths G=(V,E) , (s1,t1), β¦, (sk,tk) and need hi edge-disjoint paths per pair Want paths for each pair to be βshortβ: no more than d edges in total Minimize congestion
19
Decompose f(e,i) into multi-route hi-route-flow.
Min C for each i flow of value hi from si to ti f(e,i) β€ 1 for all e, i βi f(e,i) β€ C for all e f(e,i) β₯ 0 for all e, I βe f(e,i) β€ d for all i Decompose f(e,i) into multi-route hi-route-flow. Apply filtering to find flow of value at least Β½ on paths which use at most 2d edges Choose one βpath collectionβ for each (si,ti) independently from short path collections Congestion of O(log d/log log d) via LLL
20
Concluding Remarks Multiroute flows are useful to know
Several explicit and implicit applications in routing and network design β see refs in paper
21
Thank You!
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.