Presentation is loading. Please wait.

Presentation is loading. Please wait.

Dynamic Multi-resolution Data Dissemination in Storage-centric Wireless Sensor Networks Hongbo Luo; Guoliang Xing; Minming Li; Xiaohua Jia Department of.

Similar presentations


Presentation on theme: "Dynamic Multi-resolution Data Dissemination in Storage-centric Wireless Sensor Networks Hongbo Luo; Guoliang Xing; Minming Li; Xiaohua Jia Department of."— Presentation transcript:

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

3 3 Storage-centric Sensor Nets  Many applications are data-intensive [Ganesan03]  Structure health monitoring  Accelerometer@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  16~512 MB/sensor is recently demoed  Answer user queries on demand  Each storage node creates a data dissemination tree

4 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 5 Why Are Data Rates Important  Data rate determines total power cost  Radio power cost varies in different states  TX: 21.2~106.8 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 6 Agenda  Storage-centric wireless sensor networks  Formulation of multi-resolution data dissemination  Online tree construction and adaptation  Performance evaluation  Conclusions

7 7 An Example of Minimizing Total Radio Power  a sends to c at normalized rate of r = data rate/bandwidth  Two network configurations  a →c, b sleeps a → b → ca → b → c  Assumptions  Only source and relay nodes remain active  a→c has the worst quality  c(a,c) > c(a,b) and c(b,c)  c(x,y) is expected num of TXs from node x to y a c b

8 8 Average Power Consumption a b c a’s avg. power c’s avg. power  Configuration 1: a → c, b sleeps  Configuration 2: a → b → c θ (a,c) θ (b,c) θ (a,b) z z z

9 9 Optimal Network Configuration Transmission power dominates: use short and reliable links Idle power dominates: use long (but lossier) links since more nodes can sleep 3z 2z Power Consumption r0r0 1

10 10 Modeling Broadcast Advantage source t 1, r 1 t 2, r 2 u  Considering both u  t 1 and u  t 2  z is only counted once  Take the max of r i θ(u,t i ) for all sinks θ(u,v 1 ) θ (u,v 2 )  Considering u  s 1 only

11 11 Min-power Multi-resolution Data Dissemination (MMDD)  Given traffic demands I={(t i, r i )} and G(V,E), find a tree T(V´, E´) minimizing  zV|'|  Sleep scheduling + power-aware multicast  MMDD is NP-Hard node cost, independent of data rate d(u): set of decedents of u c(u): set of children of u

12 12 Agenda  Storage-centric wireless sensor networks  Formulation of multi-resolution data dissemination  Online tree construction and adaptation  Performance evaluation  Conclusions

13 13 Online Incremental Tree Algorithm  When a new sink t with rate r comes  Assign each edge (u,v) a cost  z+r θ(u,v), if (u,v) not on existing tree  (r θ(u,v) - max r i θ(u,v i )) +, 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 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 15 Path Quality Estimation with Increased Data Rate  Y l and Y h are min power from s to t under r l and r h  Found under cost metric z+r θ(u,v)  Theorem I: If the r l drops to r h, then power cost of Y l is no more than the min power under r h by:  Significance: path quality degradation can be estimated solely by known information all symbols are known!

16 16 Path Quality Estimation with Increased Data Rate  Theorem II: If r l increases to r h, then power cost of Y l is no more than the min power under r h by all symbols are known!

17 17 Path-quality based Tree Adaptation  Suppose sink t i changes rate from r i to r i *  Computes ∆P, the difference between current power and the min power under r i *  If ∆P×T i > β, find a new path using r i *, otherwise, continue to use the existing path  βis the energy cost of finding a shortest path  T i is the duration of new rate r i *

18 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 [r min, r max ], the performance ratio is (r max /r min )|D|, if r min ≤ r ≤ r max holds

19 19 Reference-rate based Tree Adaptation Logic  Source keeps max, min, and avg. rates of all existing requests: r min, r max, r avg  When a new request arrives  Update r min, r max to r’ min and r’ max  If r avg not in [r’ min, r’ max ], compute new avg. rate r’ avg and find a new tree using r’ avg

20 20 Agenda  Storage-centric wireless sensor networks  Formulation of multi-resolution data dissemination  Online tree construction and adaptation  Performance evaluation  Conclusions

21 21 Simulation Environment  Prowler simulator extended by Rmase project  Prowler: http://www.isis.vanderbilt.edu/projects/nest/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 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 23 Fixed Data Rates Low-rate case: each request is randomly chosen within 0.5~2 packets per active window Mixed-rate case: 1/3 requests are randomly chosen within 20~40 packets per active window

24 24 Rate- vs. Path-based Adaptation Bursty-rate case: each request alternates bw high (120~200 pkts) and low (120~200 pkts ) rates 10 times Unknown rate duration: Each request randomly changes its rate 10 times; Duration of each rate is randomly chosen from100~1000s

25 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


Download ppt "Dynamic Multi-resolution Data Dissemination in Storage-centric Wireless Sensor Networks Hongbo Luo; Guoliang Xing; Minming Li; Xiaohua Jia Department of."

Similar presentations


Ads by Google