Heather Zheng Department of Computer Science p p University of - PowerPoint PPT Presentation

Heather Zheng Department of Computer Science p p University of California, Santa Barbara CS201 UCLA, May 22, 2008 1

� Explosion of wireless networks and devices � Static spectrum assignments are inefficient p g � Under ‐ utilization + over ‐ allocation � Artificial spectrum scarcity � Solution: Migrate from long ‐ term static spectrum � Solution: Migrate from long ‐ term static spectrum assignment to dynamic spectrum access 2

� Scalability and speed � Scalability and speed � Support a large number of nodes � Adapt to time ‐ varying demands Ad t t ti i d d � Efficiency + Fairness � Maximize spectrum utilization M i i t tili ti � Avoid conflict � Reliability R li bilit � Provide QoS Manhattan (Courtesy of Wigle.net) 3

� Introduction � Dynamic Spectrum Management � Distributed spectrum coordination for fast Distributed spectrum coordination for fast adaptation adaptation p � Interference ‐ aware admission control to provide reliability � Conclusion and ongoing work 4

G G Goal Goal : Allocate spectrum to maximize system l l Node Collaboration Node Collaboration utility Assumption Assumption : 100% willingness to collaborate Action: Iterative Explicit Coordination Action: Iterative Explicit Coordination Action: Iterative Explicit Coordination Action: Iterative Explicit Coordination • Self ‐ organize into coordination groups • Negotiate to allocate spectrum in each group • Iteratively set up groups to improve utility l l • Fast convergence: coordination stops when no local improvement can improve utility Cao & Zheng, SECON 2005, Crowncom07, JSAC08, MONET08 5

Limited neighbor coordination to Limited neighbor coordination to Isolation Isolation between reduce complexity groups • One-to-one bargaining g g • One-buyer-multi-seller bargaining Self-contained group coordination to prevent group conflicts � Restricted modifications � Isolated bargaining group Each local improvement will improve the global system utility 6

Fast Convergence : The system Fast Convergence Node Collaboration Node Collaboration converges after at most O( N 2 ) local adjustments, N = network size Guaranteed Spectrum Allocation Guaranteed Spectrum Allocation : Each Guaranteed Spectrum Allocation Guaranteed Spectrum Allocation : Each node n’s allocated spectrum A(n) ≥ Poverty Line PL (n) ⎢ ⎥ L ( n ) Total usable spectrum = PL ( n ) ⎢ ⎥ + + 1 ⎣ ⎣ ⎦ ⎦ D D ( n n ( ) ) 1 Conflict degree Cao & Zheng, SECON 2005 7

Perc Perc centage centage of Insta of Insta ances ances P Percentage ge of instanc nces 10% 15% 20% 5% 5% 0 0 1 1 2 2 Node utility / lower bound A(n)/PL(n) A(n)/PL(n) 3 3 4 4 5 5 8

� Each channel i has a weight of B i (n) � Each channel i has a weight of B (n) � Each node’s spectrum allocation A(n)= ∑ a i (n)B i (n) ( ) ∑ i ( ) i ( ) � Extended poverty line A(n) > PL(n) = ∑ ∑ B B i ( n n ( ) ) i − i PL n Max B n ( ) ( ) + i d n ( ) 1 i Cao & Zheng, Crowncom07 9

� Each infrastructure node n supports t n users f � Maximize end ‐ user fairness � Each infrastructure node’s spectrum has a h f d h lower bound ⎛ ⎞ ⎢ ⎥ ⎜ ⎜ ⎟ ⎟ ⎢ ⎢ ⎥ ⎥ M M > ⋅ ⎜ − ⎟ A ( n ) t 1 ⎢ ⎥ ∑ + n ⎜ ⎟ t t ⎜ ⎜ ⎟ ⎟ ⎢ ⎢ ⎥ ⎥ n k ⎣ ⎣ ⎦ ⎦ ⎝ ⎝ ⎠ ⎠ = k N ( n ) 10

11

� Use poverty line to initiate coordination Use po ert line to initiate coordination � Enable multiple parallel coordination events � Minimize adaptation delay Minimize adaptation delay 12

B Bargaining ends; i i d Bargaining timer expires Enabled Enabled Bargaining Bargaining Send request; Receive request; Request Disable ACK Receiver timer expires ACK/disable message Disable Disabled We can also regulate the coordination format to avoid disabling neighbors 13

# of Local coordination scales Adaptation delay flattens out p y linearly with the # of APs because of parallelism. 1Mbps Wireless Backhaul running CSMA/CA among APs 14

Minimum disturbance to Quick adaptation to local Quick adaptation to local neighbors dynamics 15

Centralized, topology based Centralized, topology based 1. 1. optimization based on Modified Graph Coloring Decentralized device coordination 2. 2a. Explicit coordination through negotiation ‐ based local coordination 2b. Implicit coordination through rule ‐ based independent adjustment Focus on AP based scenarios Determine how many and which Determine how many and which � channels to use at each AP Manhattan (Courtesy of Wigle.net) 16

outage Spectrum Channels time Unreliable spectrum access How can we regulate nodes’ spectrum demand to maintain • Improved utilization reliability? 17 17

� Introduction Introd ction � Dynamic Spectrum Management � Distributed spectrum coordination for fast adaptation Di ib d di i f f d i � Interference ‐ aware statistical admission control to provide reliability id li bilit � Conclusion and Future Work 18

Admission based on statistical traffic information … … time D namic spectr m allocation Dynamic spectrum allocation based on short ‐ term demand 19 Cao & Zheng, INFOCOM08 19

− γ Guarantee: Prob(total traffic exceeds C)< e 5 4 C f γ 3 ( , s ) 2 Router capacity = C 1 0 20

Stricter than the original interference constraints Original interference constraints g Simplified linear interference constraints p 21

� Our strategy: Uniform S gy � Based on the following analytical observations � Uniform S is optimal under uniform traffic statistics � Under non ‐ uniform traffic statistics, the use of uniform S has bounded degradation g(s) � Under non ‐ uniform traffic statistics, the optimal uniform S is bounded 22

No Adm Outage = 60% SPARTA O tage < 2% Outage < 2% Peak Rate Outage =0% g � SPARTA almost doubles utilization from PRA � SPARTA loses up to 40% utilization but greatly reduces � SPARTA loses up to 40% utilization but greatly reduces outage rate compared to no admission control case 23

� Issues in Dynamic Spectrum Systems y p y Deployment � Utilization, fairness � Reliability, scalability Theoretical models � Our contributions : Collaborative � Our contributions : Collaborative Security Incentives spectrum sharing for large ‐ scale networks � Distributed coordination for fast Routing system convergence � Rule regulated self ‐ adjustment for � Rule regulated self ‐ adjustment for Spectrum allocation p simple deployment Device coordination � Providing reliability and efficiency via statistical admission control Hardware: software radios � Much more to be done 24