Enabling Dynamic Spectrum Allocation in Cognitive Radio Networks
Enabling Dynamic Spectrum Allocation in Cognitive Radio Networks Yuan Advisor: Dr. William Arbaugh Co-Advisor: Dr. Ashok Agrawala Ph. D Defense Computer Science Department University of Maryland, College Park Sept. 17, 2007
Introduction & Motivation Spectrum Shortage Unlicensed bands become over-crowed!
Introduction & Motivation Spectrum Shortage No Spectrum Available to Be Allocated in Spectrum Allocation Table from 30 MHz to 30 GHz in US
Introduction & Motivation Artificial Spectrum Shortage Dissertation Focus: 5% Tackle spectrum shortage problem by improving spectrum 30 MHz ~ 30 GHz Source: Shared Spectrum Company utilization and efficiency
Limitations of Existing Spectrum Allocation Methods Command & Control In US, FCC controls how to allocate the spectrum Extremely unbalanced spectrum utilization Fixed Channelization Further divide spectrum into fixed channels of equal bandwidth Limit network capacity ◦ TV bands: 15% in 2004* and cause unfairness ◦ Over-crowed unlicensed across a network band Source: FCC released date sheet
Dissertation Overview Dynamic Spectrum Allocation The key idea ◦ Actively detect unused spectrum ◦ Dynamically create suitable # of channels Maximize spectrum utilization Improve spectrum efficiency, minimize interference ◦ Adaptively adjust channel bandwidth Consider local user/traffic distribution New hardware support: Cognitive Radio ◦ Unused band detection ◦ Reconfigurable radio parameters
Dissertation Overview Apply Concept of Dynamic Spectrum Allocation Exploit White Spaces in TV bands A Improve Spectrum Efficiency in WLANs complete hardware- software system, dynamic channelization structure ◦ Accommodates # of neighboring APs KNOWS ◦ Reliably detects unused freq ◦ Efficiently shares spectrum b-SMART: A a distributed ◦ Allocates bandwidth considering user distribution A scalable MAC design dynamic spectrum ◦ Handles various user populations allocation algorithm ◦ Exploits rate diversity
Dissertation Overview Key Contribution The concept of Dynamic Spectrum Allocation ◦ KNOWS exploits white spaces in licensed bands Remarkable 200% throughtput improvement as compared with fixed allocation schemes ◦ Dynamic channels improve spectrum efficiency in unlicensed bands Significantly improve system throughput and fairness in WLANs
Outline Introduction & Motivation Dissertation Overview KNOWS System Design and Evaluation New Channelization Structure for WLANs Conclusions
KNOWS: Problem Formulation Resource Goal Robust White Space Detection White Space Support Data Networking in the TV bands Cognitive Radios • Sensing • Reconfigurabilit y Functionality Features Simple, distributed Efficient, practical Dynamic Access to White Space
White Spaces in TV bands Open 51 TV channel 21 - ◦ 512 MHz ~ 698 MHz Unlicensed in 2009 Low frequency band -60 Primary users Dynamic Fragmented Uneven size “White spaces” dbm -100 470 MHz Frequency 750 MHz
KNOWS Design Overview Physical-layer Capability MAC-layer Function Spectrum sensing Collaborative sensing ◦ Every 30 min required by FCC Parallelism & connectivity Highly reconfigurable Adaptive bandwidth ◦ Frequency, bandwidth, power Design highlights How many transceivers? ◦ CMAC: ◦ ONE ◦ Based on a control channel Design highlights ◦ Spectrum Allocation Table ◦ One scanner/receiver ◦ One transceiver ◦ b-SMART: distributed dynamic spectrum allocation algorithm Yuan, Paramvir Bahl, Ranveer Chandra, Philip A. Chou, John Ian Ferrell, Thomas Moscibroda, Srihari Narlanka and Yunnan Wu, KNOWS: Kognitive Networking Over White Spaces. Proceedings of IEEE Dy. SPAN, Dublin, Ireland (April 17 -20, 2007)
Hardware Platform Scanner/Receiver Scan: 400 MHz ~ 928 MHz o Recv: 900 ISM band, 5 MHz o Reconfigurable transceiver Can dynamically adjust channel-width and centerfrequency from 400 MHz to 928 MHz o Contiguous 5, 10, 20, 40 MHz o Power control o Transceiver can tune to contiguous spectrum bands only! Frequency
KNOWS Architecture
KNOWS Architecture Next!
Basic Idea in CMAC + b-SMART • CMAC • Spectrum Allocation Table • b-SMART o. Records Time-Spectrum Blocks (f, ∆f, t, ∆t) Freq Time Spectrum Block TV bands tf f Control Channel ISM band t Time Spectrum Block tt Time Spectrum Block Time Spectrum Block Time Spectrum Block time Yuan, Paramvir Bahl, Ranveer Chandra, Thomas Moscibroda, Srihari Narlanka and Yunnan Wu, Allocating Dynamic Time-Spectrum Blocks in Cognitive Radio Networks. Proceedings of ACM Mobi. Hoc, Montreal, Canada 2007
CMAC Overview Sender Receiver RTS ◦ Indicates intention for transmitting ◦ Contains suggestions for available time-spectrum block (b-SMART) CTS DTS Waiting Time CTS DTS ◦ Data Transmission re. Servation ◦ Announces reserved time-spectrum block to neighbors of sender t+∆t DATA ACK Time-Spectrum Block ◦ Spectrum selection (received-based) t ◦ Annouces selected time-spectrum block (f, ∆f, t, ∆t)
Spectrum Allocation Table Nodes record info for reserved time-spectrum blocks TV Frequency Primary Users Control channel IEEE 802. 11 -like Congestion resolution The above depicts a possible Spectrum Allocation Table 1) Primary users (fragmentation) 2) In multi-hop neighbors have different views Time
b-SMART Which time-spectrum block should be reserved…? ◦ (f, ∆f, t, ∆t) How long…? How wide…? b-SMART Design (distributed spectrum allocation over white spaces) Principles 1. Try to assign each flow blocks of bandwidth B/N B: Total available spectrum N: Number of disjoint flows 2. Choose optimal transmission duration t Long blocks: Higher delay Short blocks: More congestion on control channel
b-SMART Upper bound Tmax~10 ms on maximum block duration Nodes always try to send for Tmax 1. Find bandwidth ∆f, for which the time used to out packets in current queue is closest to Tmax 2. If ∆f ≥ then ∆f : = 3. Find placement of (∆f, ∆t) block that minimizes finishing time and non-overlap with any other block 4. If no such block can be placed due to prohibited bands then ∆f : = ∆f /2 ∆f ∆f = Tmax Spectrum Allocation Table
KNOWS: Performance Evaluation • Simulate in Qual. Net • Total 80 MHz, 1 MHz to 1. 2 Mbps • Bandwidth: 5, 10, 20, 40 MHz • Control channel: 5 MHz • Switch overhead: 50 µs • Backlogged UDP flows, and TCP flows
KNOWS in Single Hop Network Aggregate Throughput of Disjoint UDP flows 90 Throughput (Mbps) 80 70 60 50 40 30 20 10 0 0 2 4 6 8 10 12 14 16 18 # of flows KNOWS significantlyb-SMART outperforms systems based on fixed allocations! b-SMART in in Contiguous Spectrum Fragmented Spectrum SSCH: Multichannel MAC Single TV channel
KNOWS in Chain Network 20 b-SMART in Contiguous Spectrum Throughput (Mbps) 18 16 14 b-SMART in Fragmented Spectrum 12 10 8 SSCH: Multichannel MAC 6 4 Single TV channel 2 0 0 2 4 6 8 10 12 14 16 18 20 # nodes in the chain KNOWS significantly improve throughput and reduces interferences!
KNOWS Summary KNOWS: hardware-software system ◦ Detect unoccupied frequencies in licensed TV bands ◦ Support dynamic spectrum allocation using b. SMART Performance Evaluation ◦ Analysis results match simulation results ◦ Quantify the effect of fragmentation, traffic type, application type, multiple-hop network, routing protocols, mobility
Outline Introduction & Motivation Dissertation Overview KNOWS System Design and Evaluation New Channelization Structure for WLANs Conclusions
Fixed Channels in WLANs 22 MH z 1 2 3 4 2400 MHz 5 7 AP usage in WLANs tends to User populations served by APs fluctuate considerably 9 10 11 2483. 5 MHz Limitations of Fixed Channels Limit Network Capacity ◦ be unbalanced 8 IEEE 802. 11 b Unbalanced Traffic Distribution 6 # of neighboring APs is small Cause Interference ◦ # of neighboring APs is large
Dynamic Channelization Structure The key idea: ◦ Dynamically create suitable # of channels Accommodate # of neighboring APs ◦ Adaptively adjust channel bandwidth Consider user/traffic distribution Yuan, Paramvir Bahl, Ranveer Chandra, Thomas Moscibroda and Yunnan Wu, Un. Channelize the Channels in WLANs. Proceedings of ACM Mobi. Com Poster, Montreal, Canada 2007
# collisions per client Throughput (Mbps) Simulation Study in Large Scale Offices Average number of interfering APs Qualnet Settings: • IBM trace data, 50 APs • 1000 m x 1000 m flat • 80 MHz spectrum, • Switch overhead: 50 us • 1 MHz -> 1. 2 Mbps • Greedy. Raising: enables dynamic channels • Ra. C: channel selection algorithm based on fixed channels
Outline Introduction & Motivation Dissertation Overview KNOWS System Design and Evaluation New Channelization Structure for WLANs Conclusions
Spectrum Shortage Problem A critical problem to solve to support fast growth of wireless technologies Key Contribution Dynamic Spectrum Allocation significantly improves spectrum utilization and efficiency ◦ KNOWS exploits white spaces in licensed bands ◦ Dynamic channels improve spectrum efficiency in unlicensed bands
Future Work Deploy KNOWS system, measure performance and further improve the design Further study performance of dynamic channels Apply Dynamic Spectrum Allocation in market-based approach ◦ Maximize revenue ◦ Reduce interference
Enabling Dynamic Spectrum Allocation in Cognitive Radio Networks Yuan Ph. D defense QUESTIONS?
- Slides: 32