Description
Medium access control (MAC) is a fundamental problem in wireless networks.

In ad-hoc wireless networks especially, many of the performance and scaling issues

these networks face can be attributed to their use of the core IEEE 802.11 MAC

protocol: distributed coordination function (DCF).

Medium access control (MAC) is a fundamental problem in wireless networks.

In ad-hoc wireless networks especially, many of the performance and scaling issues

these networks face can be attributed to their use of the core IEEE 802.11 MAC

protocol: distributed coordination function (DCF). Smoothed Airtime Linear Tuning

(SALT) is a new contention window tuning algorithm proposed to address some of the

deficiencies of DCF in 802.11 ad-hoc networks. SALT works alongside a new user level

and optimized implementation of REACT, a distributed resource allocation protocol,

to ensure that each node secures the amount of airtime allocated to it by REACT.

The algorithm accomplishes that by tuning the contention window size parameter

that is part of the 802.11 backoff process. SALT converges more tightly on airtime

allocations than a contention window tuning algorithm from previous work and this

increases fairness in transmission opportunities and reduces jitter more than either

802.11 DCF or the other tuning algorithm. REACT and SALT were also extended

to the multi-hop flow scenario with the introduction of a new airtime reservation

algorithm. With a reservation in place multi-hop TCP throughput actually increased

when running SALT and REACT as compared to 802.11 DCF, and the combination of

protocols still managed to maintain its fairness and jitter advantages. All experiments

were performed on a wireless testbed, not in simulation.
Reuse Permissions
  • Downloads
    PDF (1.5 MB)

    Details

    Title
    • Smoothed Airtime Linear Tuning and Optimized REACT with Multi-hop Extensions
    Contributors
    Date Created
    2018
    Subjects
    Resource Type
  • Text
  • Collections this item is in
    Note
    • Masters Thesis Computer Science 2018

    Machine-readable links