1

Data-Centric Storage in Sensornets with GHT, A Geographic Hash Table

Sylvia Ratnasamy, Scott Shenker,

Brad Karp, Ramesh Govindan, Deborah Estrin,

Li Yin and Fang Yu

ICSI/UCB/USC/UCLA

2

Outline

Background Existing Schemes Data-Centric Storage Conclusion

3

Background

Sensornet♦ A distributed sensing network comprised of a large number of small sensing devices equipped with

• processor • memory • radio♦ Great volume of data

Data Dissemination Algorithm ♦ Scalable ♦ Self-organizing ♦ Energy efficient

4

Observations/Events/Queries Observation

♦ Low-level output from sensors♦ E.g. detailed temperature and pressure readings

Event♦ Constellations of low-level observations♦ E.g. elephant-sighting, fire, intruder

Query♦ Used to elicit the event information from sensornets♦ E.g. locations of fires in the network

Images of intruders detected

5

Existing Schemes

External Storage (ES) Local Storage (LS) Data-Centric Storage (DCS)

6

External Storage (ES)

7

Local Storage (LS)

EventData

EventData

8

Local Storage (LS)

EventData

EventData

9

Data-Centric Storage (DCS) Events are named with keys DCS provides (key, value) pair DCS supports two operations:

♦ Put (k, v) stores v ( the observed data ) according to the key k, the name of the data♦ Get (k) retrieves whatever value is stored associated with key k

Hash function♦ Hash a key k into geographic coordinates♦ Put() and Get() operations on the same key k hash k to the same location

10

DCS – Example

(11, 28)Put(“elephant”, data)

(11,28)=Hash(“elephant”)

PDA

11

DCS – Example

PDA

(11, 28)

(11,28)=Hash(“elephant”)

Get(“elephant”)

12

DCS – Example – contd..

PDA

elephant

fire

13

Comparison Study

Metrics♦ Total Messages

• total packets sent in the sensor network

♦ Hotspot Messages• maximal number of packets sent by any particular

node

14

Comparison Study - contd.. Assume ♦ n is the number of nodes

♦ Asymptotic costs of O(n) for floods

O(n 1/2) for point-to-point routing

O(n1/2)0O(n 1/2)Cost for Storage

O(n1/2)O(n1/2)0Cost for Response

O(n1/2)O(n)0Cost for Query

DSLSES

15

Comparison Study -contd..

Dtotal, the total number of events detected Q , the number of event types queries for Dq, the number of detected events of event types

No more than one query for each event type, so there are Q queries in total.

Assume hotspot occurs on packets sending to the access point.

16

Comparison Study – contd..

Hotspot

Total

DCSLSES

nDtotal nDQn q nDnDnQ qtotal

totalDqDQ qDQ

DCS is preferable if Sensor network is large Dtotal >> max[Dq, Q]

)(summarynQnDnQ total

)(2 summaryQ

17

Geographic Hash Table (GHT)

Builds on

♦ Peer-to-peer Lookup Systems

♦ Greedy Perimeter Stateless Routing

GHT

GPSRPeer-to-peer lookup system

18

Review GPSR

Greedy forwarding algorithm Perimeter forwarding algorithm

GHT Home node

to be the node geographically nearest the destination coordinates of the packet

Home perimeter the entire perimeter that encloses the destionation.

20

Problems

Not robust enough♦ Nodes could move (new home node?)

♦ Home nodes could fail Not scalable

♦ Home nodes could become communication bottleneck

♦ Storage capacity of home nodes

21

Solutions

Perimeter Refresh Protocol♦ Extension for robustness

♦ Handles nodes failure and topology change Structured Replication

♦ Extension for scalability

♦ Load balance

22

Perimeter Refresh Protocol

PRP stores a copy of a key-value pair at each node on the home perimeter.

PRP generates refresh packets periodically.

23

Structured Replication

Use a hierarchical decomposition of the key space.

For a given root r and a given hierarchy depth d, one can compute 4d-1 mirror images of r

Simulation

Success rate the mean over all queries of the fraction of

events returned in each response, divided by the total number of events known to have been stored in the network for that key.

f the fraction of nodes that remain up for the

entire simulation.

Simulation

Stable and Static Nodes

Simulation

Static but Failing Nodes

27

Simulation

System parameters:N, the number of nodes in the systemT, the number of event types, T = 100Q, the number of event types queried forD

i, the number of detected events of event type

i. Di = 100

28

Simulation

Three version of DCSNormal DCS (N-DCS): a query returns a

separate message for each detected eventSummarized DCS (S-DCS): A query returns a

single message regardless of the number of detected events

Structured Replication DCS (SR-DCS)

29

Simulation

Test 1: Varying Q

30

Simulation

Test 1: Varying Q

31

Simulation

Test 2: Varying n

32

Simulation

Test 2: Varying n

33

Conclusion Advantages:

In DCS, relevant data are stored by name at nodes within the sensornets.

To ensure robustness and scalability, DCS uses Perimeter Refresh Protocol (PRP) and Structured Replication (SR).

Compared with ES and LS, DCS is preferable in large sensornet.

34

Conclusion

Disadvantages:GHT requires approximate knowledge of a

sensornet's boundariesOnly supports binary events, not range queries.

35

Questions?

Thanks