Data Storage Schemas in Sensor Networks

1
Data Storage Schemas in Sensor Networks

• Presenter Chengdu Huang
• 09-16-2002

2
Data-Centric Storage in Sensornets S.
Ratnasamy, D. Estrin, R. Govindan, B. Karp,
S. Shenker, L. Yin and F. Yu (2002) GPSR Greedy
Perimeter Stateless Routing for Wireless Network
B. Karp and H. Kung (Mobicom2000)
3
Outline

• Background
• Network model assumptions
• Existing schemas
• Data-centric storage
• Extensions
• Performance
• Discussion

4
Data in a Sensor Network

• What kind of data reside in sensor networks?
• Data Storage
• What, where and how?
• Others (how long, consistency etc.)
• Subject to resource limitations
• Communication bandwidth
• Power consumption
• Computation capability

5
Network Model Assumptions

• Randomly, evenly spread out
• Fairly dense
• Location service
• Stationary and stable
• Nodes are peers

6
Observations/Events/Queries

• Observation
• Low-level output from sensors
• Event
• Constellations of low-level observations
• E.g. fire, intruder
• Clients use Queries to elicit event information
  from sensor network
• E.g. Locations of fires in the network
• E.g. Images of intruders detected

7
Possible Approaches

• External Storage (ES)
• Local Storage (LS)
• Data-Centric Storage (DS)
• Directed Diffusion
• Geographically Targeted

8
External Storage (ES)
Base station
9
ES Problems
10
Local Storage (LS)
11
Local Storage (LS)
12
Data-Centric Storage (DCS)

• Data-Centric data are named
• Event data are stored, by name, at some home
  nodes.
• Queries also go to the home nodes instead of
  the nodes detected events

13
The Big Picture

• Based on geographic routing (Karp) and P2P lookup
  algorithm (Ratnasamy)

14
Distributed Hash Table (DHT)

• void Put(key, value)
• Stores value to home node in the sensor
  networks according to key
• Value Get(key)
• Retrieve value from home node in the sensor
  networks according to key

15
DCS Example Revisit
(11, 28)
(11,28)Hash(elephant)
16
DCS Example
Get(elephant)
(11, 28)
(11,28)Hash(elephant)
17
DCS Example
elephant
fire
18
Properties of DHT

• Distributed Hash Function
• Known to everybody
• Every home node takes care of roughly the same
  amount of event types
• Evenly distributed geographically
• Candidate Message Digest Algorithms
• Such as SHA-1, MD5

19
DHT - Example

• Example

Elephant
(11, 28)
MD5
Mapping to Area
5a76e813d6a0a40548b91acc11557bd2
20
GPSR

• Location service required
• Send(value, x, y)
• Nodes know identifications and positions of their
  neighbors
• Greedy forwarding
• Packets are greedily forwarded to neighbor
  closest to destination coordinates
• Perimeter forwarding

21
GPSR Greedy Forwarding
22
GPSR - Void
23
GPSR Perimeter Forwarding
Right Hand Rule Each node to receive a packet
forwards the packet to the next link
counterclockwise about itself from the ingress
link
2
X
Z
3
1
Y
24
GPSR Perimeter Forwarding
Right Hand Rule Each node to receive a packet
forwards the packet to the next link
counterclockwise about itself from the ingress
link
25
Comparison Study

• Metrics
• Total Messages
• total packets sent in the sensor network
• Hotspot Messages
• maximal number of packets sent by any particular
  node

26
Comparison Study contd

• DCS is preferable if
• Sensor network is large
• Dtotal gtgt maxDq, Q
• Summaries are used

27
Problems with DCS

• Not robust enough
• Home nodes could fail
• Nodes could move (new home node?)
• Not scalable
• Home nodes could become communication bottleneck
• Storage capacity of home nodes

28
Solutions

• Perimeter Refresh Protocol
• Extension for robustness
• Handles nodes failure and topology change
• Structured Replication
• Extension for scalability
• Load balance

29
Perimeter Refresh Protocol
(replica)
(replica)
E
D

• Key stored at location L.
• Home node A.
• Replicas D and E on the home perimeter

L
F
A
(home)
C
B
30
Perimeter Refresh Protocol
(replica)
(replica)
E
D

• Some time after node A fails, replica D
  initiates a fresh for L

L
F
C
B
31
Perimeter Refresh Protocol
(replica)
(replica)
E

• Node F becomes the new home node
• Node F recruits replicas B, C, D and E

D
L
F
(home)
C
(replica)
B
(replica)
32
Structured Replication

• Home node -gt Root (4d-1) mirror images for a
  given hierarchy depth
• Storage cost reduces
• O(N 1/2) -gt O(N 1/2/2d)
• Query cost increases
• O(N 1/2) -gt O(2dN 1/2)

33
Performance

• Metrics
• Total Usage total packets sent in the sensor
  network
• Hotspot Usage maximal number of packets sent by
  any particular node
• Map to energy consumption

34
Performance contd
Total Messages, varying queries
35
Performance contd
Hotspot Messages, varying queries
36
Performance contd
Total Messages, varying network size
37
Discussion

• Related Works
• Distributed database approach
• DataSpace
• COUGAR
• Data-centric routing/in-network aggregation
• Directed Diffusion
• Building Efficient Wireless Sensor Networks with
  Low-level Naming (SOSP 2001)
• Modeling Data-Centric Routing in Wireless Sensor
  Networks (INFOCOM 2002)
• Tiny AGgregation

38
Discussion contd

• Critiques
• Data consistency in storages
• Oversimplified summary
• Tradeoff of extensions not evaluated
• Ignored event data size as a factor impacts
  performance

39
Our Extension

• Modified Schema
• Instead of sending the raw event data to home
  nodes, nodes just send links to themselves to the
  home nodes
• Clients contact home node to get the locations of
  events first, then contact the nodes and get the
  event data there
• Tradeoff
• Total amount of data sent through the the network
  is greatly reduced
• Another indirection

40
Our extension - Performance