Distributed HopbyHop Rate Adjustment for Congestion Control in Sensor Networks

1
Distributed Hop-by-Hop Rate Adjustment for
Congestion Control in Sensor Networks

• Presented by
• Manmohan Voniyadka
• Sapna Dixit
• Vipul Bhasin
• Vishal Kumar Singh

2
Agenda

• Problem description
• Solution requirements
• Algorithm description
• Congestion pricing scheme
• Rate adjustment using price of congestion
• Protocol details
• Open Issues
• Schedules and Work Distribution

3
Problem Description

• Rate Adjustment to avoid congestion
• Calculation of congestion pricing via a
  distributed mechanism

4
Solution Requirements

• Develop a congestion pricing scheme
• Develop a relationship between rate and the
  congestion pricing.
• Modify CODA Protocol to incorporate these changes

5
Congestion Pricing Scheme

• Factors for pricing
• MAC utilization.
• Congestion Price at Downstream Nodes.
• Calculate a cumulative congestion pricing factor.
• Calculate pricing factor in a distributed manner.
• Weigh the pricing on congestion nodes
  appropriately.

6
Congestion Pricing Calculation

• Cumulative Congestion Pricing
• wi relative weight of downstream node, i
• wi
• N is the Nth node from the congested node
• cpi congestion pricing at the ith node
• pj is a marking function (As given in 1)

7
Rate Adjustment Using Congestion Pricing

• Use an AIMD Strategy

8
Protocol Details

• Conditions for Backpressure Origination
• Send backpressure when threshold is exceeded as
  done in CODA
• Threshold based on channel sampling
• Header Format Changes
• Extra field for price of congestion value in
  backpressure messages
• End-to-End Loop Control

9
Open Issues

• Relative weight in Congestion pricing is only
  based on congestion price at downstream nodes,
  Need to add Transmission rate at ith node as
  factor in weight.
• Whether to use price of congestion for end to end
  loop control for controlling the source rate
  during persistent congestion.
• What happens with Asymmetric links ?

10
Changes to existing CODA

• Changes in Suppression Message generation.
• Threshold based on MAC Utilization / channel
  Sampling.
• Calculate and Send Congestion Price.
• Changes in ReceiveBackPressure Message -gt
  AdjustSourceRate.

11
Glossary

• CPj P (j) MAC Utilization (j).
• CPj is the congestion price for using node j
  per unit time.
• P(j) is the marking function at node j and
  determines the fraction of flow to be marked.
• MAC Util is the fraction of time node j spends
  in receiving and re-transmitting to next hop.
• P (j) P (j) (y - tj) / y
• y is sum of MAC utilization time by all flows at
  node j.
• tj is a parameter for controlling MAC time
  utilization.
• P (j) indicates fraction of flow exceeding the
  threshold parameter. If link quality is bad , We
  make tj ltlt 1 for lower MAC time utilization.

12
References

• 1 Chieh-Yih Wan, Shane B. Eisenman and Andrew
  T. Campbell, CODA Congestion Detection and
  Avoidance in Sensor Networks, ACM SenSys 2003,
  November 2003.
• 2 Y. Yi and S. Shakkottai. Hop-by-hop
  Congestion Control over a Wireless Multi-hop
  Network, IEEE INFOCOM, 2004.