Throughput and Fairness in A Hybrid Channel Access Scheme for Ad Hoc Networks

1
Throughput and Fairness in A Hybrid Channel
Access Scheme for Ad Hoc Networks

• Yu Wang and J. J. Garcia-Luna-Aceves
• Computer Communication Research Group
• Department of Computer Engineering
• University of California at Santa Cruz, U.S.A.
• URL http//www.cse.ucsc.edu/ccrg

2
Outline

• Introduction
• The hybrid channel access scheme
• Simulation results
• Conclusion and future work

3
Introduction

• Medium access control (MAC) for ad hoc networks
• Distributed, no centralized control
• Throughput scarce channel resource
• Fairness to avoid severe throughput degradation
  experienced by some unlucky nodes
• IEEE 802.11 is the de facto standard, but a lot
  of work has been and is still being done.

4
Introduction

• Throughput
• Collision avoidance (CA) is important to
  alleviate the hidden terminal problem.
• CA usually includes a four-way handshake
  (RTS-CTS-data-ACK) between a pair of sending and
  receiving nodes sender-initiated (SI).
• Receiver-initiated CA schemes have also been
  proposed a node polls its neighbors actively to
  see if they have packets for itself.
• Example Receiver-initiated multiple access
  (RIMA) protocols.

5
Introduction

• Throughput (contd)
• Receiver-initiated (RI) CA
• Rationale Contention is mainly at the receivers
  side (data packet lasts longer) and receiver has
  better knowledge of the contention around itself
  Less control packet overhead.
• Caveat A good traffic estimator and an
  appropriate polling discipline is mandatory.
• Fairness problem
• Location dependent contention
• Binary exponential backoff (BEB) in IEEE 802.11
  aggravates the problem because it favors the node
  that last succeeds.

6
Introduction

• Fairness example node 1 faces more contention
  than node 3.

0
1
2
3
0
RTS0
1
RTS2
RTS2
2
RTS2
RTS2
3
T1
T2
7
Introduction

• Existing approaches to achieve fairness
• Category 1 Max-Min fairness
• Reduce the ratio between max throughput and min
  throughput of flows, at either a nodes level or
  a flows level.
• Category 2 Fair queuing (FQ) approach
• Adapt wireline FQ disciplines to ad hoc networks.
• Flow contention graph as a useful abstraction
• Flows are tagged as either leading or lagging,
  and then backoff window is adjusted accordingly.
• Remarks tradeoff between throughput and
  fairness SI channel access.

8
Introduction

• A hybrid scheme that combines SI and RI can be
  more effective
• It distributes the burden of initiating CA
  handshake between two nodes.
• Better fairness can be expected.
• Designed to be compatible with IEEE 802.11
• Nodes not using the hybrid scheme are not
  affected.
• Only known frame types are used, so CTS reused.
• Correct collision avoidance (like RIMA) is not
  enforced, hence throughput enhancement is not the
  goal.

9
The Hybrid Channel Access Scheme

• A node can be in SI mode or RI mode.
• In SI mode, a node just uses the usual
  RTS-CTS-data-ACK handshake.
• In RI mode, a node uses CTS-data-ACK handshake,
  in which CTS serves as polling packet to the
  polled node for data packet.
• RI mode requires closer cooperation between a
  sender and a receiver.
• RI mode is indicated by an used bit in the IEEE
  802.11 frame structure.

10
The Hybrid Scheme

• Illustration of the IEEE 802.11 frame structure

11
The Hybrid Scheme

• Sender side
• After some RTS transmissions without CTS from the
  receiver, it sets RI flag in all subsequent RTS
  packets and enters RI Setup mode.
• If CTS is received, then RI-association with the
  receiver is set up and the sender will not send
  any more RTS until queue empty.

12
The Hybrid Scheme

• Receiver side
• When RTS with RI flag set is received, it
  generates CTS in its MAC queue and initiates
  CTS-data-ACK handshake in due time.

13
The Hybrid Scheme

• Receiver side (contd)
• RI-response packet (in fact, self-initiated CTS)
  is added to its MAC queue only when the
  head-of-line (HOL) packet is not an RI-response
  packet for the same polled node.
• Avoid aggravation of the fairness problem
• RI-response packets are treated like RTS packets
  for normal data packets.
• Avoid excessive delay or deadlock
• Chances for neighbor nodes to initiate CA
  handshake

14
Simulation Environments

• GloMoSim 2.0 is the network simulator.
• Two competing flows in networks with 2-4 nodes.
• 2Mbps channel with direct sequence spread
  spectrum (DSSS) parameters

15
Simulation Environments
0
0
1
2
3-4
1
0
1
2
3
4-1
2
0
1
2
3
3-3
4-2
0
1
2
3
4-3
0
1
0
1
0
1
3
2
2
2
4-8
4-10
3
4-7
3
Sample network configurations
16
Simulation Environments

• Simulation Set 1
• Two competing constant bit rate (CBR) flows
• UDP is the transport layer, thus only MAC level
  ACKs are sent.
• Simulation Set 2
• Two competing FTP flows
• FTP client sends data packet to FTP server
  without backward control traffic from the server
  other than TCP level ACKs.
• The client sends the next packet only after the
  previous one is acknowledged.

17
Simulation Results

• Simulation Set 1 with the original 802.11

(CBR, throughput in bps)
18
Simulation Results
Throughput Comparison in Different Network
Configurations (CBR, throughput in bps)
19
Simulation Results

• Simulation Set 2 with the original 802.11

(FTP, throughput in bps)
20
Simulation Results
Throughput Comparison in Different Network
Configurations (FTP, throughput in bps)
21
Simulation Results

• For CBR traffic, the fairness problem exists in
  two configurations which can be alleviated by the
  hybrid scheme.
• For FTP traffic, the fairness problem is much
  more severe and only a few cases can be
  alleviated.
• Flow control and congestion avoidance in TCP get
  in the way to achieve better fairness.
• More information exchange among nodes is
  necessary.

22
Conclusion and Future Work

• The new hybrid scheme can alleviate the fairness
  problem with almost no degradation in throughput
  and maintains compatibility with the existing
  IEEE 802.11.
• It still cannot solve the fairness problem
  conclusively, and more information exchange among
  nodes is mandatory.
• Integration with other proposed FQ schemes will
  be addressed in future work.

23
Selected References

• J. J. Garcia-Luna-Aceves and A. Tzamaloukas,
  Receiver-initiated Collision Avoidance in
  Wireless Networks,'' ACM Wireless Networks, vol.
  8, pp. 249--263, 2002.
• T. Ozugur et al., Balanced Media Access Methods
  for Wireless Networks,'' Proc. of ACM/IEEE
  MobiCom '98, pp. 21--32, Oct. 1998.
• T. Nandagopal et al., Achieving MAC Layer
  Fairness in Wireless Packet Networks,'' ACM
  MobiCom 2000, (Boston, MA, U.S.), Aug. 2000.
• N. H. Vaidya et al. Distributed Fair Scheduling
  in a Wireless LAN,'' ACM MobiCom 2000, (Boston,
  MA, U.S.), Aug. 2000.
• H. Luo et al., A New Model for Packet
  Scheduling in Multihop Wireless Networks,'' ACM
  MobiCom 2000, (Boston, MA, U.S.), Aug. 2000.
• B. Bensaou et al., Fair Medium Access in 802.11
  Based Wireless Ad-Hoc Networks,'' IEEE/ACM
  MobiHoc Workshop, Aug. 2000.
• X. Huang, On Max-min Fairness and Scheduling in
  Wireless Ad-Hoc Networks Analytical Framework
  and Implementation,'' ACM MobiHoc '01, 2001.
• H. Luo and S. Lu, A Topology-Independent Fair
  Queueing Model in Ad Hoc Wireless Networks,''
  IEEE ICNP 2000, (Osaka, Japan), Nov. 2000.
• H. Luo et al., A Self-Coordinating Approach to
  Distributed Fair Queueing in Ad Hoc Wireless
  Networks,'' IEEE INFOCOM 2001, Apr. 2001.