Databases Reference
In-Depth Information
An environment where sensors are embedded in smart objects and living
infrastructures, such as roads and buildings, provides seamless monitoring of
our living conditions and more effective means of conducting our day-to-day
activities. Consider the use of sensor networks that are attached to roads for
tra
c monitoring. In a peak period when many vehicles pass a particular
intersection with tra
c lights, the sensors will communicate with the tra
c
light system to change the system settings to allow more vehicles to pass
through on a congested route. Indirectly, such systems will improve the quality
of life for people.
The state of miniaturization and cost of production of wireless sensors make
it quite feasible to build such systems. However, one major impediment pre-
vents their implementation. The data arising from the sensors would be in-
cessant, and due to the rapid generation of sensory data from the nodes, the
data will rapidly become extremely large. Therefore, any such system must
continue to keep up with data using superior filtering, storage, and processing.
To fully understand the information gained from these sensor data, a mech-
anism to learn and adapt to its characteristics is required. The sensor data
that fit into the spatial and/or temporal domains would generate a signature
and/or trend for a given space and/or time. This signature can be detected
using a pattern recognition approach.
The data deluge phenomenon that we are currently experiencing is affecting
the way that we process these data. It is not realistic to use simple analytical
means to understand the information obtained from a multitude of sources
with large magnitudes of size and dimension. A paradigm shift in data process-
ing is necessary. Common pattern recognition applications that execute within
a CPU-centric environment to recall or memorize megabytes or gigabytes of
data may not be effective when processing terabytes of data. The recognition
scheme must be considered from a larger perspective, i.e., the Internet-scale
perspective.
pattern recognition is a term that will be referred to as recognition involv-
ing large-scale data. These data may be coming from different sources ranging
from sophisticated devices and facilities to simple but large-scale data collec-
tion mechanisms, such as a wireless sensor network (WSN) deployment for
monitoring environmental conditions. The key aspect of Internet-scale pat-
tern recognition is a larger capacity for the recognition of data. Therefore,
the means by which data are analyzed must scale with the growth of the
data. There are several key technologies that contribute to this Internet-scale
pattern recognition approach. These include distributed systems, parallel com-
puting technologies, and machine intelligence. In its simplest definition, a
distributed system is a collection of independent computers that appear to
be a single coherent system to its users. Formally, a distributed system is a
computer architecture containing multiple inter-connected processors. These
processors are inter-connected via communication networks that perform a
particular task or operate collectively.
