Wireless Sensor Networks : Technology, Protocols, and Applications

EXISTING MIDDLEWARE 255
reasonable response times [8.2]. IrisNet is not designed specifically for resourcelimited
WSNs. For example, IrisNet has not considered localized algorithms or possible
WSN application features.
8.4.3 AMF (Adaptive Middleware Framework)
The adaptive middleware framework (AMF) proposed in [8.3] exploits ‘‘resource
and application QoS trade-offs’’ and ‘‘predictability of sensor readings’’ to reduce
the energy consumed in the process of information collection. The assumption is
that it is possible to collect approximate data at predetermined accuracy levels
while satisfying an application’s QoS. AMF has ‘‘sensor-side’’ and ‘‘server-side’’
components which bridge the application layer with the underlying sensor network
infrastructure. It supports both precision- and prediction-based adaptation. Serverside
components include application quality, data quality requirement translation,
adaptive precision setting, sensor selection, sensor data management, and fault tolerance.
Sensor-side components include sensor-state management and precisiondriven
adaptation. AMF has an energy-efficient message-updating mode, where
the sensor sends an update to the server only when the measurement value exceeds
the previous value or the value predicted beyond a given error level [8.3]. The server
maintains a list of active sensors (active list) and a list of historic values for each
sensor over a specified time period. To support prediction-based adaptations, a sensor
and the server store a set of prediction models and choose the best one according
to the network status. AMF attempts to trade off between resource and quality
during information collection. In this context it reduces sampling frequency without
compromising the accuracy of results [8.5].
8.4.4 DSWare (Data Service Middleware)
DSWare [8.4] resides between the application and network layers, integrates various
real-time data services, and provides a database-like abstraction to applications. It
includes several components: data storage, data caching, group management, event
detection, data subscription, and scheduling. In DSWare [8.4], data are replicated in
multiple physical nodes mapped to a single logical node using a hash-based mapping.
Queries are directed to any of the nodes to avoid collision and to balance the
load among the nodes. A data caching service in DSWare monitors the current use
of copies and determines whether to increase or reduce the number of copies and
whether to move some copies to another location by exchanging information in the
neighborhood [8.4]. DSWare incorporates group management to provide localized
cooperation among sensor nodes and to perform a global objective. It also performs
real-time scheduling for queries in WSN. A data subscription service in DSWare
minimizes communication among sensor nodes.
DSWare provides a novel event-detection mechanism that is reliable and energy
efficient. As described earlier, a compound event is assumed to include subevents
that may be correlated, and its occurrence can be measured by a confidence
function. The result of the confidence function is called confidence. When the
confidence is greater than a threshold minimal confidence, a compound event is