Presentation is loading. Please wait.

Presentation is loading. Please wait.

Metrics for Characterizing QoS in Multi-Channel, Multi-Radio,Multi-Rate Wireless Meshes Ranjan Pal.

Similar presentations


Presentation on theme: "Metrics for Characterizing QoS in Multi-Channel, Multi-Radio,Multi-Rate Wireless Meshes Ranjan Pal."— Presentation transcript:

1 Metrics for Characterizing QoS in Multi-Channel, Multi-Radio,Multi-Rate Wireless Meshes
Ranjan Pal

2 Introduction

3 Basics of Wireless Mesh Networks
Nodes may be fixed or mobile Interconnected with wireless links User devices act as both clients and routers Targeted towards civilian applications Example: MIT Roofnet

4 Wired Internet Backbone
AP AP WR WR WR User Cluster User Cluster

5 Specialized Features of Wireless Mesh Networks
Nodes equipped with multiple radios Can transmit on orthogonal multiple channels Quality of channels vary over time (SNR) - Shadowing & Fading Radios adjust to transmit at multiple rates

6 Network Model

7 dda cbd D B daf A F C E Flow Split dcb

8 Problem Definitions

9 Problem 1 Given a demand vector D, find the expected multicommodity flow achieved i.e Compute l e <+ such that expected multicommodity flow is l¢D

10 Compute l e <+ such that expected multicommodity flow is l¢D
Problem 2 Given a demand vector D and cost limit CL, find the expected multicommodity flow achieved i.e Compute l e <+ such that expected multicommodity flow is l¢D

11 Problem 3 For a given multicommodity flow demand D, rank the links of the network in order of importance in achieving the demand i.e The link whose variation in capacity by a small amount, affects the achievability of D the maximum, is ranked 1 and so on…

12 Motivation

13 Why QoS metrics?? flows in MR2-MC WMN’s is an open issue
Evaluating QoS(analytically) of multicommodity flows in MR2-MC WMN’s is an open issue [Kodialam et.al ’05], [Mishra et.al ’05] Judging suitability of applications for a n/w Budgeting analysis for a service company Performance tuning for bettering QoS support

14 My Contributions

15 Developed a polynomial-time algorithm to
evaluate l for a non cost constrained n/w evaluate l for a cost constrained n/w Used Fussell-Vesely* and Birnbaum* methods to characterize link importances and have shown thru simulations that using FV is better

16 Outline of Solution Methodologies

17 Remove “Curse of Dimensionality” Prove algorithms in ‘P’
Apply AcQoS(D) Apply AcQoS(D,CL) Prove algorithms in ‘P’ Use FV and Birnbaum methods to find link importances

18 Some facts about the algorithms
Stochastic flow based (obey Ford-Fulkerson) Requires the values of pi’ for each link Basic idea is to find a set of minimal capacity vectors under which the demand is satisfied and the probability that any system capacity vector is greater than the minimals Algorithms shown to be polynomial (in P) using combinatorial analysis

19 Link Importance Measures

20 Birnbaum Measures Gives the link that most significantly affects the achievability of a given D Two types – SAD and MAD SAD (Average Sum of Absolute Deviations) - Considers the possible state levels of a link MAD (Mean/Expected Absolute Deviation) - Considers possible state levels & prob. of a link being in that state

21 Fussell-Vesely Measures
Relative measures that account for the average change in achieved QoS when link states “negatively” contribute to achieved QoS Two Types – GFVM and MFVM GFVM (General Fussell-Vesely Measure) - Considers the possible state levels of a link MFVM (Mean Fussell-Vesely Measure) - Considers possible state levels & prob. of a link being in that state

22 Simulation Results

23 Topology

24 Performance With Three Channels

25 6 5 4 2 1 3

26 General Birnbaum and Fussell-Vesely

27 Mean Birnbaum and Fussell-Vesely

28 Observations on Link Importances
Variances are very low (very robust) Ranks for general Birnbaum and FV differ Ranks for Mean Birnbaum and FV are same Mean measures more consistent and robust Prefer Mean FV as its variance is the lowest

29 Developed ways to characterize QoS in multi-radio
Conclusions Developed ways to characterize QoS in multi-radio multi-channel, multi-rate wireless meshes Used Birnbaum and Fussell-Vesely measures to find link importances w.r.t to a flow demand Future Work Compute bounds of achieved QoS in MR2-MC wireless mesh networks Tackle statistical link dependencies

30 Questions ???


Download ppt "Metrics for Characterizing QoS in Multi-Channel, Multi-Radio,Multi-Rate Wireless Meshes Ranjan Pal."

Similar presentations


Ads by Google