Data EveryWhere: Dealing Data From Sensors N/W

Data,Data Everywhere:
Technical Seminar Presentation

Dealing Data
From Sensors Network

Under the guidance of :

Mr.Indraneel Mukhopadhyay
Technical Seminar Adviser

Rashmi Rekha Das 1

Data EveryWhere: Dealing Data From Sensors N/W

Introduction
Technical Seminar Presentation

What is sensor network?

It is continuously collecting and communicating the
data string to the database.
What is its real life application?
Monitoring physical phenomena such as climate,
concentration of contaminants, building structure and
response to earthquakes, etc.
How does it operate?
It will operate as black-boxes that record diagnostic
data, performance data, history of the object, etc., e.g.,
in vehicles, cell phones and computers.

Rashmi Rekha Das 2

Data EveryWhere: Dealing Data From Sensors N/W

Introduction
Technical Seminar Presentation

What is the work of sensor network?

Data gathering techniques using network primitives,
data cleaning techniques such as cleaning and
querying of noisy data, discovering outliers, and
handling incomplete data due to missing values and
handling of erroneous sensors.

Rashmi Rekha Das 3

Data EveryWhere: Dealing Data From Sensors N/W

Background
Technical Seminar Presentation

Limitations
Limited battery life, storage and processing capabilities
which enforce special data handling algorithms and
architectures for sensor data streams, that explicitly
incorporate these resource constraints.
Packet loss and topology changes.
Capabilities
Sensors are capable of constructing and maintaining a
locally consistent view of its neighborhood .

Rashmi Rekha Das 4

Data EveryWhere: Dealing Data From Sensors N/W

Background
Technical Seminar Presentation

Sensor streaming Vs Traditional streaming

Sensor Streaming Traditional Streaming
Samples of population Entire population
Imprecise and noisy Exact and error-free
Moderate size Huge data
Expensive No cost
Continuous domain Frequent items,distinct

Rashmi Rekha Das 5

Data EveryWhere: Dealing Data From Sensors N/W

In-network aggregation
Technical Seminar Presentation

Two approaches
Tiny Aggregation (TAG) service for adhoc sensor network.
Aggregation for monitoring wireless sensor networks .
Query Evaluation
Distribution phase
The query is distributed to every node in the network. A
tree rooted at the base-parents.
Collection phase
Parents collect data from children at specific time intervals.
Rashmi Rekha Das 6
Data EveryWhere: Dealing Data From Sensors N/W

In-network aggregation
Technical Seminar Presentation

Routing
Ability to deliver query requests to every node.
Ability to provide one or more routes from every node to the
root. In particular, it uses a tree-based routing where one
node is the root.
Performance
TAG dramatically decreases communication and yields an
order of magnitude reduction in communication cost
compared to a centralized approach.
Heavy packet loss and link asymmetry can be quite common
in sensor networks.

Rashmi Rekha Das 7

Data EveryWhere: Dealing Data From Sensors N/W

In-network storage
Technical Seminar Presentation

What is the storage approach in sensor networks?

There are generally three approaches for storing data in
sensor networks: external, local, and data centric storage
approaches .
External storage, sensor data is continuously sent to a
“powerful” collecting point.
Local storage, data are stored locally at its original node.
Data centric storage (DCS), data are stored by name, and therefore,
there is a cost of communicating the data from its original node to
the node where it will be stored. However, cost of querying in this
case is reduced since queries are directed to the node that stores the
data and not to every node in the network.
Rashmi Rekha Das 8
Data EveryWhere: Dealing Data From Sensors N/W

Architecture for sensor streams

Technical Seminar Presentation

These systems enable three types of queries on sensor

streams: historical queries, snapshot queries, and long-
running queries.
An architecture is proposed for managing multiple queries
over streams of sensors and traditional data sources.

Rashmi Rekha Das 9

Data EveryWhere: Dealing Data From Sensors N/W

Fjording sensor streams

Technical Seminar Presentation

The architecture consists of two major components: Fjord and sensor.

Users issue queries using sensor catalogs to the query processor.

The sensors relay their readings to the proxy either as raw or

Processed signals, e.g., aggregate.

The proxy packs these samples as tuples and forwards them to the

query processor. The processor processes the query using the

provided data and, finally, output the answer to the end users.

Rashmi Rekha Das 10

Data EveryWhere: Dealing Data From Sensors N/W

Erroneous sensor data

Technical Seminar Presentation

Online cleaning
The uncertainty model can be computed either at the sensor level

or at the database server when sensor data arrives .

As sensor data flows, we incorporate the likelihood of obtaining

the observed values with prior knowledge of the distribution of

the true values and, consequently, we obtain posterior uncertainty

models of the unknown true values.

Rashmi Rekha Das 11

Data EveryWhere: Dealing Data From Sensors N/W

Conclusion
In particular, we surveyed current research in storage
Technical Seminar Presentation

and gathering of sensor data, architectures for querying

sensor streams, and data cleaning.
Also, imprecision due to noise, missing values,
massiveness of data, etc., and deriving new optimization
metrics in query evaluation.
A decision making infrastructure will be built on top of
these sensors. This infrastructure will enable extraction
of any needed information, directly from sensors, on
demand, and in real-time.
Rashmi Rekha Das 12
Data EveryWhere: Dealing Data From Sensors N/W
Technical Seminar Presentation

THANK YOU

Rashmi Rekha Das 13