Dynamic Multiresolution Data Dissemination in Storagecentric Wireless Sensor Networks

1
Dynamic Multi-resolution Data Dissemination in
Storage-centric Wireless Sensor Networks
Hongbo Luo Guoliang Xing Minming Li Xiaohua
Jia Department of Computer Science City
University of Hong Kong
2
Agenda

• Storage-centric wireless sensor networks
• Formulation of multi-resolution data
  dissemination
• Online tree construction and adaptation
• Performance evaluation
• Conclusions

3
Storage-centric Sensor Nets

• Many applications are data-intensive Ganesan03
• Structure health monitoring
• Accelerometer_at_100Hz, 30 min/day, 80Gb/year
• Micro-climate and habitat monitoring
• Acoustic video, 10 min/day, 1Gb/year
• Store most data in network
• Storage has low cost and power consumption
• 16512 MB/sensor is recently demoed
• Answer user queries on demand
• Each storage node creates a data dissemination
  tree

4
Dynamic Multi-resolution Data Dissemination

• Requests have different temporal resolutions
• "report temperature readings every 1 minute"
• "report light readings every 2 minutes"
• Requests are dynamic
• New requests can arrive anytime
• Data rates of existing requests can change
• Optimal dissemination tree is not fixed!

5
Why Are Data Rates Important

• Data rate determines total power cost
• Radio power cost varies in different states
• TX 21.2106.8 mW, RX and idle 32 mW, Sleeping
  0.001 mW
• Total energy cost is sum of power in each state
  weighted by the working time
• Exploring diversity of rates reduces power due to
  broadcast wireless channel

6
Agenda

• Storage-centric wireless sensor networks
• Formulation of multi-resolution data
  dissemination
• Online tree construction and adaptation
• Performance evaluation
• Conclusions

7
An Example of Minimizing Total Radio Power
c

• a sends to c at normalized rate of
• r data rate/bandwidth
• Two network configurations
• a ?c, b sleeps
• a ? b ? c
• Assumptions
• Only source and relay nodes remain active
• a?c has the worst quality
• c(a,c) gt c(a,b) and c(b,c)
• c(x,y) is expected num of TXs from node x to y

b
a
8
Average Power Consumption
c

• Configuration 1 a ? c, b sleeps

z
?(b,c)
cs avg. power
as avg. power
z
?(a,c)
b
?(a,b)
z
a

• Configuration 2 a ? b ? c

9
Optimal Network Configuration
Transmission power dominates use short and
reliable links
Power Consumption
3z
2z
1
r0
Idle power dominates use long (but lossier)
links since more nodes can sleep
10
Modeling Broadcast Advantage

• Considering u?s1 only

source
u

• Considering both u?t1 and u?t2
• z is only counted once
• Take the max of ri?(u,ti) for all sinks

?(u,v1)
?(u,v2)
t1, r1
t2, r2
11
Min-power Multi-resolution Data Dissemination
(MMDD)

• Given traffic demands I(ti , ri ) and G(V,E),
  find a tree T(V, E) minimizing

node cost, independent of data rate
d(u) set of decedents of u c(u) set of children
of u

• Sleep scheduling power-aware multicast
• MMDD is NP-Hard

12
Agenda

• Storage-centric wireless sensor networks
• Formulation of multi-resolution data
  dissemination
• Online tree construction and adaptation
• Performance evaluation
• Conclusions

13
Online Incremental Tree Algorithm

• When a new sink t with rate r comes
• Assign each edge (u,v) a cost
• zr ?(u,v), if (u,v) not on existing tree
• (r ?(u,v) - max ri?(u,vi)), otherwise
• Find the shortest path from source to t
• Theorem total power cost D times of power
  cost of optimal tree found offline
• D is num of requests arrived so far

14
Lightweight Tree Adaptation

• When data rates of existing requests change
• Power efficiency of a tree degrades
• Constructing a new tree is expensive
• Path-quality based tree adaptation
• Monitor the quality of each path
• Find a new path if quality drops below a
  threshold
• Reference-rate based tree adaptation
• Monitor the reference of all data rates
• Find a new tree if reference exceeds a threshold

15
Path Quality Estimation with Increased Data Rate

• Yl and Yh are min power from s to t under rl and
  rh
• Found under cost metric zr ?(u,v)
• Theorem I If the rl drops to rh, then power
  cost of Yl is no more than the min power under rh
  by
• Significance path quality degradation can be
  estimated solely by known information

all symbols are known!
16
Path Quality Estimation with Increased Data Rate

• Theorem II If rl increases to rh, then power
  cost of Yl is no more than the min power under rh
  by

all symbols are known!
17
Path-quality based Tree Adaptation

• Suppose sink ti changes rate from ri to ri
• Computes ?P, the difference between current power
  and the min power under ri
• If ?PTi gt ß, find a new path using ri,
  otherwise, continue to use the existing path
• ßis the energy cost of finding a shortest path
• Ti is the duration of new rate ri

18
Reference-rate based Tree Adaptation

• Find paths using same rate r for all sinks
• Significantly reduces the overhead
• Theorem for a set of requests D with rates in
  rmin, rmax, the performance ratio is
  (rmax/rmin)D, if rmin r rmax holds

19
Reference-rate based Tree Adaptation Logic

• Source keeps max, min, and avg. rates of all
  existing requests rmin, rmax, ravg
• When a new request arrives
• Update rmin, rmax to rmin and rmax
• If ravg not in rmin, rmax, compute new avg.
  rate ravg and find a new tree using ravg

20
Agenda

• Storage-centric wireless sensor networks
• Formulation of multi-resolution data
  dissemination
• Online tree construction and adaptation
• Performance evaluation
• Conclusions

21
Simulation Environment

• Prowler simulator extended by Rmase project
• Prowler http//www.isis.vanderbilt.edu/projects/n
  est/prowler/
• Rmase http//www2.parc.com/spl/projects/era/nest/
  Rmase/
• Implemented USC model Zuniga et al. 04 to
  simulate lossy links of Mica2 motes
• 40 Kbps bandwidth, transmission power of 11.6 mA,
  idle power of 8 mA
• Routing nodes keep active 50s in every 500s
• Simulated different workload patterns
• High, low, mixed, busty data rates

22
Simulation I Fixed Data Rates

• Three baseline algorithms
• Min transmission count tree (MTT)
• Shortest-path tree of expected of TXs
• Transmission count Steiner tree (TST)
• Approx. min Steiner tree of expected of TXs
• Similar to power-aware multicast algorithms
• Data rate Steiner tree (DST)
• Approx. min Steiner tree based on data rates
• Similar to data dissemination algorithm SEAD
  kim03

23
Fixed Data Rates
Low-rate case each request is randomly chosen
within 0.52 packets per active window
Mixed-rate case 1/3 requests are randomly chosen
within 2040 packets per active window
24
Rate- vs. Path-based Adaptation
Bursty-rate case each request alternates bw high
(120200 pkts) and low (120200 pkts) rates 10
times
Unknown rate duration Each request randomly
changes its rate 10 times Duration of each rate
is randomly chosen from1001000s
25
Conclusions

• Multi-resolution data dissemination
• Models all states of radio, link quality, data
  rates, broadcast advantage
• An online tree construction algorithm
• Handles dynamic arrivals of data requests
• Two lightweight tree adaptation heuristics
• Maintain power-efficiency under dynamic rates