Database Reference
In-Depth Information
database system. The method allows for composite event detection, and
the declarative language used (SNLog) is a variant of Datalog.
3. Event Processing over RFID Streams
One essential goal for RFID applications is to map objects and their
behaviors in the physical world into the virtual counterparts and their
virtual behaviors in the applications by semantically interpreting and
transforming RFID data. Application logic can often be devised and
engineered as complex RFID events, and once such complex events are
detected, the semantics can be automatically interpreted. Based on the
purposes of RFID data processing, RFID applications can be generally
classified as two categories: i) history-oriented object tracking supported
through temporal database or data warehousing based solutions [75, 76],
and ii) real-time oriented monitoring and stream processing through
complex RFID event processing techniques. Complex RFID event pro-
cessing plays a critical role on interpreting the semantics of RFID data
and supporting real-time monitoring applications.
Basic theory of complex event processing has been intensively studied
in the area of active database. There exist several processing models,
including automata-based, Petri net- based, matching tree- based and
directed graph- based. As these processing models did not fully consider
the characteristics and complex semantics of RFID events, they can not
be applied to RFID complex event processing immediately.
Different from the events in traditional active databases [54, 55] and
message-based processing systems [56], RFID events have their own
unique characteristics. First, RFID events are temporally constrained:
both the temporal distance between two events and the interval of a
single event are critical for the event detection. In addition, RFID ap-
plications can also generate non-spontaneous events - events that cannot
detect their occurrences by themselves unless they either get expired or
are explicitly queried. These include negated events (an event which
does not occur) and temporal constrained events, for example, an event
that occurs within a certain period. Such non-spontaneous events pose
new challenges for event processing. Moreover, the actions from RFID
events are quite different: they are normally database updates and mes-
sages, and neither trigger new primitive events for the system, nor lead
to a cascade of rule firings as in active databases.
Some large-scale IT application providers [57-60] and academic in-
stitutions [61] had provided many platforms to collect data from RFID
readers and pump the collected data to upper down-streaming systems.
However, these platforms currently only support simple event process-
