+ All Categories
Home > Documents > 1 Caching/storage problems and solutions in wireless sensor network Bin Tang CSE 658 Seminar on...

1 Caching/storage problems and solutions in wireless sensor network Bin Tang CSE 658 Seminar on...

Date post: 20-Dec-2015
Category:
View: 217 times
Download: 4 times
Share this document with a friend
22
1 Caching/storage problems and solutions in wireless sensor network Bin Tang CSE 658 Seminar on Wireless and Mobile Networking
Transcript

1

Caching/storage problems and solutions in wireless sensor network

Bin Tang

CSE 658 Seminar on Wireless and Mobile Networking

2

Outline Introduction Web caching and

replication/placement Caching/storage in sensor network (based on Mobisys’ 03 paper) Our approach and preliminary

simulation results

3

Motivation of data placement/ caching in sensor network

Collection and delivery sensory data

Power conservation of each node is important

Communication cost is dominant among sensing, processing and communication cost.

Save communication cost - caching

4

Problem and objective Given network topology and user

access pattern, decide the number and location of web content replicas

Objective function can be minimized

clients’ latency, bandwidth, or overall cost function

5

General Approach

Map the placement problem onto a graph optimization problem Facility location problem Minimum K-median problem

Solve graph optimization problem Using various approximation algorithms

6

Minimum K-Median Problem

Given a complete graph G=(V,E), d(j), c(i,j)

d(j): # demands c(i,j): distance

between node i and j

Find a subset V’ V with |V’| = K s.t. it minimizes vV minwV’ d(v)c(v,w)

Facility Location problem Given a complete graph

G=(V,E), f(i), d(j), c(i,j) f(i): # cost of building i d(j): # demands c(i,j): distance between

node i and j Find a subset V’ V s.t. it

minimizes f(i) + vV minwV’

d(v)c(v,w)

7

Algorithms Tree based algorithm

underlying topologies are trees, and model it as a dynamic programming problem

O(N3M2) for choosing M replicas among N potential places

Random Pick the best among several random

assignments Hot spot

Place replicas near the clients that generate the largest load

8

Greedy algorithm Calculate costs of assigning clients to replicas Select replica with lowest cost

Super-Optimal algorithm Lagrangian relaxation + subgradient method

9

Data Storage/placement in sensor network What’s the difference with web caching?

High density makes topology more flexible Caching service runs on every sensor node

Approaches: Data-centric

Attribute-based naming Data stored by name

Web caching approach (heuristics) “Energy-conserving Data Placement and

Asynchronous Multicast in Wireless Sensor Networks” by Sagnik Bhattacharya, et al. (Mobisys’03)

10

Model Multiple observers(sinks) Focus locales, representatives Publish-subscribe model

Problem formulation – construction of a minimum-cost Steiner tree, which connects sensor node to observers

11

MS(teiner)T, MS(spanning)T, MK-Median problem.

Data placement/tree construction Join the multicast tree Copy creation and migration Leave the multicast tree

Simulation – comparison with unicast model,

synchronous multicast, directed diffusion

12

Our approach Minimum Steiner tree:

N={1, 2, …, n}, S is source node with D dist(i,j), hop numbers b/t i and j U: update frequency of D a(i): access frequency of node i to D Find a set of intermediate nodes M N, to

minimize: U*(optimal cost of a Steiner tree covering S UM) +iN minjM a(i)dist(i,j)

O(N6M2) for choosing M replicas among N potential places in tree topology.

13

Simplified model Steiner tree cost is replaced by

individual unicast cost

Update cost is considered as constraint, objective function is

t(N, M) = iN minjM a(i)dist(i,j)

14

Greedy algorithm Benefit of A: B(A, M)=t(N,M)-t(N,M U A) Greedy algorithm: In each iteration,

select the node with maximum B until the update constraint is reached.

Optimal bound for both single and multiple documents storage of each node: 1-1/e

15

Simulations

Simulation model: A network of (200 <= N <= 1000)

sensor nodes 100m x 100m area Transmission radius (12m <=R <=

25m)

16

Update cost constraint

Total enegy dissipated vs. maximum update cost

0

250

500

750

1000

0 50 100 150 200

Maximum update cost (num of hops)

En

ger

y d

issi

pat

d

(nu

mb

er o

f h

op

s)

17

Energy saved from greedy algorithm

Engergy saved vs. number of nodes

0

1000

2000

3000

0 250 500 750 1000

Number of nodes

En

erg

y sa

ved

(n

um

o

f h

op

s)

18

The effect of update cost

Total energy dissipated plotted against sensor update reate

0

200

400

0 1 2 3 4 5 6 7

Average update rate/average request rate

En

erg

y d

issi

pat

ed

19

The role of transmission rate

Total Energy dissipated vs. transmission radius

0

50

100

150

200

0 5 10 15 20 25 30

Transmission radius

En

ger

gy

dis

sip

ated

(n

um

o

f h

op

s)

20

Future work Multiple documents stored in

sensor node

Distributed dynamic caching scheme.

21

References “Energy-conserving Data Placement and

Asynchronous Multicast in Wireless Sensor Networks” (Mobisys’03)

Sagnik Bhattacharya et al. On the placement of Web Server Repilcas”

(INFOCOM’01) Lili Qiu et al.

“Data-Centric Storage in Sensornets” (WSNA’02) Deborah Estrin et al.

22

Thanks!


Recommended