MACNetwork CrossLayer approaches to Improve Performance of Wireless Multihop Networks

1
MAC-Network Cross-Layer approaches to Improve
Performance of Wireless Multi-hop Networks

• Jerry Le
• 06239420
• 24 FEB 2007

2
Preface

• Traditional TCP/IP built on wired networks
• Differences in wireless networks
• - no links, but circles
• (the shared nature of wireless channel)
• - higher packet loss rate and
• collision probability
• - multihop, each node may be a
• router itself

3
Preface

• An example

- by fair MAC schedule node 2 gets B/3, node 1
gets B/2 - node 1 is a hidden terminal to node
3 - may lead to packet loss and collision! -
solution MAC should share some information with
Network layer
4
Preface

• Whats not covered
• - security issues
• - power saving
• - ...
• Focus
• - The cross-layer approaches between MAC and
  Network to address the most fundamental issues
  (e.g. throughput, delay, QoS)
• - Why, How, and How good (bad)
• Variations of cross-layer
• - combining functions of MAC and Network
  1,2,3,5
• - 2.5 Layer approach 4,6
• - Radically new architecture 7

5
Outline

• Background
• - Characteristics of wireless multihop network
• Scheduling optimization
• Admission control and Rate control
• Radically new approach
• Comments Conclusion

6
BackgroundCharacteristics of wireless multi-hop
network

• Important distances
• - Transmission/Reception
• range
• - Carrier Sensing Range
• (Interference range)
• Multi-hop routing
• - Each node can serve
• as a router
• Capacity dynamics
• - detector errors due to mobility, multi-path,
  fading, etc.
• - packet loss due to collisions
• - rate auto-adaptation

7
Outline

• Background
• Scheduling optimization
• - intraflow and interflow contention1
• - slot-based scheduling2
• Admission control and Rate control
• Radically new approach
• Comments Conclusion

8
Scheduling optimizationIntraflow and Interflow
contention 1

• Intraflow contention
• - contention between upstream and downstream
  nodes along one path
• - may lead to overflow and packet loss
• - solution assign high channel access priority
  to each node when it just receives a packet,
  achieving optimal scheduling for chain topology
  (B/3)

1 H. Zhai and Y. Fang, Distributed flow control
and medium access in multihop ad hoc networks,
IEEE Transactions on Mobile Computing, vol. 5,
no. 11, 2006.
9
Scheduling optimizationIntraflow and Interflow
contention 1

• Intraflow contention
• - solution in more details
• Option 1 assign higher priority to packets
  that have traversed more hops (needs Network
  Layer information)
• Option 2 set the backoff window size to be
  the minimum whenever a node receives a packet to
  forward

1 H. Zhai and Y. Fang, Distributed flow control
and medium access in multihop ad hoc networks,
IEEE Transactions on Mobile Computing, vol. 5,
no. 11, 2006.
10
Scheduling optimizationIntraflow and Interflow
contention 1

• Interflow contention
• - contention between the transmissions in
  several paths
• - solution backward-pressure scheduling
• block the upstream nodes when the downstream
  nodes have enough packets to send
• resume transmission of upstream nodes when
  transmission finished

Packets may accumulate and get lost at node 1!
1 H. Zhai and Y. Fang, Distributed flow control
and medium access in multihop ad hoc networks,
IEEE Transactions on Mobile Computing, vol. 5,
no. 11, 2006.
11
Scheduling optimizationIntraflow and Interflow
contention 1

• Interflow contention
• - solution in more details
• modify RTS-CTS exchange in 802.11 MAC

1 H. Zhai and Y. Fang, Distributed flow control
and medium access in multihop ad hoc networks,
IEEE Transactions on Mobile Computing, vol. 5,
no. 11, 2006.
12
Scheduling optimizationIntraflow and Interflow
contention 1

• Per-flow fairness
• - number of packets of one flow in the queue can
  not exceed source-flow threshold
• - Round-Robin scheduling among flows

1 H. Zhai and Y. Fang, Distributed flow control
and medium access in multihop ad hoc networks,
IEEE Transactions on Mobile Computing, vol. 5,
no. 11, 2006.
13
Scheduling optimizationSlot-based scheduling2

• Overlay MAC 2
• - A Layer 2.5 approach
• - TDMA-like scheduling
• - Weighted Slot Allocation (WSA), assigning a
  weight (priority) to each node
• - loose synchronization required
• - not suitable for RT (Real-Time) traffic (slot
  size too big)
• - implementation

2 A. Rao and I. Stoica, An overlay mac layer
for 802.11 networks," in Proceedings of
Mobisys'05, 2005.
14
Outline

• Background
• Scheduling optimization
• Admission control and Rate control
• - CACP3, PAC4
• - SoftMAC5, SWAN6
• Radically new approach
• Comments Conclusion

15
Admission control and Rate control CACP3,
PAC4

• CACP (Contention-aware Admission Control
  Protocol) 3
• PAC (Perceptive Admission Control) 4
• - Hop-by-Hop admission control for RT
  (Real-time) traffic, guaranteeing QoS
• - Admission decision based on
• 1) free bandwidth
• 2) bandwidth to be consumed by a new
  flow

How to evaluate?
3 Y. Yang and R. Kravets, Contention-aware
admission control for ad hoc networks,"
IEEE Transactions on Mobile Computing, vol. 4,
no. 1, 2005. 4 I. D. Chakeres and E. M.
Belding-Royer, PAC Perceptive admission control
for mobile wireless networks," in Proceedings of
QShine'04, 2004.
16
Admission control and Rate controlCACP3, PAC4

• Evaluation of free bandwidth
• - Finding Local Available bandwidth
• Option 1 Size of backoff window
• Option 2 Queue length
• Option 3 Number of collisions
• Option 4 Delay
• Option 5 Channel Busy Time

can not sense the bandwidth utilization if not
transmitting
Not enough for the evaluation of free bandwidth!
3 Y. Yang and R. Kravets, Contention-aware
admission control for ad hoc networks,"
IEEE Transactions on Mobile Computing, vol. 4,
no. 1, 2005. 4 I. D. Chakeres and E. M.
Belding-Royer, PAC Perceptive admission control
for mobile wireless networks," in Proceedings of
QShine'04, 2004.
17
Admission control and Rate controlCACP3, PAC4

• Evaluation of free bandwidth
• - Finding c-neighbor Available bandwidth
  (consider the potential
• interference to the transmissions of neighboring
  nodes in interference range)

c-neighbor available bandwidth 0.7B
3 Y. Yang and R. Kravets, Contention-aware
admission control for ad hoc networks,"
IEEE Transactions on Mobile Computing, vol. 4,
no. 1, 2005. 4 I. D. Chakeres and E. M.
Belding-Royer, PAC Perceptive admission control
for mobile wireless networks," in Proceedings of
QShine'04, 2004.
18
Admission control and Rate controlCACP3, PAC4

• Evaluation of free bandwidth
• - How to find c-neighbor available bandwidth?
• option 1 query all nodes in interference
  range of their local
• available bandwidth, then choose the minimum
  value 3
• (the query message may traverse multiple
  hops)
• option 2 use a lower carrier sensing
  threshold 4

3 Y. Yang and R. Kravets, Contention-aware
admission control for ad hoc networks,"
IEEE Transactions on Mobile Computing, vol. 4,
no. 1, 2005. 4 I. D. Chakeres and E. M.
Belding-Royer, PAC Perceptive admission control
for mobile wireless networks," in Proceedings of
QShine'04, 2004.
19
Admission control and Rate controlCACP3, PAC4

• Evaluation of the bandwidth to be consumed
• - Sum up the time for DIFS-RTS-CTS-DATA-SIFS-ACK
  
• Incorporate the local admission control into
  Multi-hop routing
• - incorporate required bandwidth into Route
  Request
• - intermediate nodes reply the Route Request
  based on its
• c-neighbor available bandwidth
• - a new path is formulated only when the
  bandwidth is
• sufficient at every node along the path

3 Y. Yang and R. Kravets, Contention-aware
admission control for ad hoc networks,"
IEEE Transactions on Mobile Computing, vol. 4,
no. 1, 2005. 4 I. D. Chakeres and E. M.
Belding-Royer, PAC Perceptive admission control
for mobile wireless networks," in Proceedings of
QShine'04, 2004.
20
Admission control and Rate controlSoftMAC5,
SWAN6

• SoftMAC 5
• - A Layer 2.5 approach
• - Similar Admission Control for RT traffic
• - Rate Control for BE (Best-Effort) traffic
• (when channel condition varies, the rate of
  BE traffic must be
• adapted such that fraction of free
  bandwidth remains the same!)
• - A new approach to evaluate free bandwidth
• FAT (Fraction of Air Time)
• - Key differences 1) take into account the
  channel dynamics 2) implementation!

5 H. Wu, Y. Liu, Q. Zhang, and Z.-L. Zhang,
Softmac Layer 2.5 collaborative mac for
multimedia support in multi-hop wireless
networks," IEEE Transactions on Mobile Computing,
vol. 6, no. 1,2007. 6 G.-S. Ahn, A. T.
Campbell, A. Veres, and L.-H. Sun, Supporting
service differentiation for real-time and
best-effort traffic in stateless wireless ad hoc
networks (swan)," IEEE Transactions on Mobile
Computing, vol. 1, no. 3, 2002.
21
Admission control and Rate controlSoftMAC5,
SWAN6

• SoftMAC 5
• - FAT in more details
• 1) Consumed FAT the fraction of time
  consumed for
• transmission in an interval
  (calculation takes into
• consideration a stable loss
  probability)
• 2) Residual FAT of a node the fraction of
  free time
• of a node
• 3) Residual FAT of a link the minimum of
  Residual
• FAT among two end nodes

5 H. Wu, Y. Liu, Q. Zhang, and Z.-L. Zhang,
Softmac Layer 2.5 collaborative mac for
multimedia support in multi-hop wireless
networks," IEEE Transactions on Mobile Computing,
vol. 6, no. 1,2007. 6 G.-S. Ahn, A. T.
Campbell, A. Veres, and L.-H. Sun, Supporting
service differentiation for real-time and
best-effort traffic in stateless wireless ad hoc
networks (swan)," IEEE Transactions on Mobile
Computing, vol. 1, no. 3, 2002.
22
Admission control and Rate controlSoftMAC5,
SWAN6

• SWAN (Stateless Wireless Ad-hoc Network) 5
• - an older approach
• - source-based admission control for RT traffic
• (estimate network utilization by probe
  message, may not be accurate)
• - MAC Layer AIMD-like rate control for BE
  traffic
• (no consideration on the impact to
  neighboring RT traffic)
• - easily beaten by CACP3

5 H. Wu, Y. Liu, Q. Zhang, and Z.-L. Zhang,
Softmac Layer 2.5 collaborative mac for
multimedia support in multi-hop wireless
networks," IEEE Transactions on Mobile Computing,
vol. 6, no. 1,2007. 6 G.-S. Ahn, A. T.
Campbell, A. Veres, and L.-H. Sun, Supporting
service differentiation for real-time and
best-effort traffic in stateless wireless ad hoc
networks (swan)," IEEE Transactions on Mobile
Computing, vol. 1, no. 3, 2002.
23
Outline

• Background
• Scheduling optimization
• Admission control and Rate control
• Radically new approach
• Comments Conclusion

24
Radically new approachA new protocol
architecture7

• Why the old architecture does not work well?
• - hop-centric operation inherited from wired
  network
• - Relaying is not a primitive function of
  Physical Layer
• - Fail to utilize the broadcast nature of
  wireless channel

the unefficient receive-store-process- queue-forwa
rd-contend-transmit process
7 R. Ramanathan, Challenges A radically new
architecture for next generation mobile ad hoc
net-works," in Proceedings of Mobicom'05, 2005.
25
Radically new approachA new protocol
architecture7

• Solutions
• - A Relay-oriented Physical Layer
• coupling receive and transmission, full duplex
• from decode-and-forward to amplify-and-forward
• incorporate routing functions

Ideally, the packet never go beyond physical
layer of the intermediate nodes
7 R. Ramanathan, Challenges A radically new
architecture for next generation mobile ad hoc
net-works," in Proceedings of Mobicom'05, 2005.
26
Radically new approachA new protocol
architecture7

• Solutions
• - Path Access Control (PAC)
• rather than one hop, PAC reserves the floor for
  multiple hops
• The RTS/CTS message may traverse multiple hops
  to reserve bandwidth

7 R. Ramanathan, Challenges A radically new
architecture for next generation mobile ad hoc
net-works," in Proceedings of Mobicom'05, 2005.
27
Radically new approachA new protocol
architecture7

• Solutions
• - Cooperative Transport (exploiting physical
  layer diversity)
• Let idle nodes help transport packets
• A packet may be transmitted by multiple nodes to
  enhance SNR
• A MIMO-like receiver is required

7 R. Ramanathan, Challenges A radically new
architecture for next generation mobile ad hoc
net-works," in Proceedings of Mobicom'05, 2005.
28
Outline

• Background
• Scheduling optimization
• Admission control and Rate control
• Radically new approach
• Comments Conclusion

29
Comments Conclusion

• A Cautionary Perspective8
• - cross-layer design does bring performance
  gain, at
• least in the demonstrated scenarios
• - many are short-term gains
• - architecture vs. performance
• (long-term) (short-term)
• - may sacrifice robustness and convenience
• - cross-layer design has no order improvement
• Interaction between different approaches
• - how to compare one with another?
• - currently, most approaches are mutually
  exclusive

8 Vikas Kawadia and P. R. Kumar, A Cautionary
Perspective on Cross Layer Design. IEEE Wireless
Communication Magazine. pp. 3-11, vol. 12, no. 1,
February 2005
30
Comments Conclusion

• New research possibility on cross-layer design
• - Network coding 9
• - Capability of multi-channel or multiple NICs
  on one
• node 10,11

9 S. Katti, H. Rahul, W. Hu, D. Katabi, M.
Medard and J. Crowcroft, XORs in the air
practical wireless network coding. In Proceedings
of Sigcomm06, 2006 10 Draves, R., Padhye, J.,
and Zill, B. Routing in multi-radio, multi-hop
wireless mesh networks. In Proceedings of ACM
Mobicom'04 (2004). 11 So, J., and Vaidya, N. H.
Routing and channel assignment in multichannel
multihop wireless networks with single network
interface. In QShine05 (2005).
31
Thanks for your attention!
32
Reference

• 1 H. Zhai and Y. Fang, Distributed flow control
  and medium access in multihop ad hoc
• networks, IEEE Transactions on Mobile Computing,
  vol. 5, no. 11, 2006.
• 2 A. Rao and I. Stoica, An overlay mac layer
  for 802.11 networks," in Proceedings
• of Mobisys'05, 2005.
• 3 Y. Yang and R. Kravets, Contention-aware
  admission control for ad hoc networks," IEEE
• Transactions on Mobile Computing, vol. 4, no. 1,
  2005.
• 4 I. D. Chakeres and E. M. Belding-Royer, PAC
  Perceptive admission control for mobile
• wireless networks," in Proceedings of QShine'04,
  2004.
• 5 H. Wu, Y. Liu, Q. Zhang, and Z.-L. Zhang,
  Softmac Layer 2.5 collaborative mac for
• multimedia support in multi-hop wireless
  networks," IEEE Transactions on Mobile Computing,
  
• vol. 6, no. 1,2007.
• 6 G.-S. Ahn, A. T. Campbell, A. Veres, and
  L.-H. Sun, Supporting service
• differentiation for real-time and best-effort
  traffic in stateless wireless ad hoc
• networks (swan)," IEEE Transactions on Mobile
  Computing, vol. 1, no. 3, 2002.
• 7 R. Ramanathan, Challenges A radically new
  architecture for next generation
• mobile ad hoc net-works," in Proceedings of
  Mobicom'05, 2005.
• 8 Vikas Kawadia and P. R. Kumar, A Cautionary
  Perspective on Cross Layer
• Design. IEEE Wireless Communication Magazine.
  pp. 3-11, vol. 12, no. 1, February
• 2005