Image Processing Reference
In-Depth Information
. Highly connected networks
. Numerous memory-constrained endsystems
. Stringent timeliness requirements
. Adaptive online reconfiguration of computation and communication policies and
mechanisms
Networkedembeddedsystemschallengeassumptionsaboutresourceavailabilityandscalemade
by classical approaches to distributed computing, and thus represent an active research area with
many open questions. For example, advances in microelectro mechanical systems (MEMS) hard-
ware technology have made it possible to move software closer to physical sensors and actuators to
make more intelligent use of their capabilities. To realize this possibility, however, new networked
embedded systems technologies are needed. For example, hardware infrastructure for such systems
may consist of a network of hundreds or even thousands of small microcontrollers, each closely
associated with local sensors and actuators.
2.1.1 Multiple Design Dimensions
The following four dimensions drive the design choices for development of many networked embed-
ded systems:
. Temporal predictability
. Distribution
. Feature richness
. Memory constraints
There is often a contra-variant relationship between some of these design forces. For example, the
left side of Figure . illustrates that feature richness may suffer when footprint is reduced. Similarly,
a real-time embedded system's temporal performance must be maintained even when more or fewer
features are supported, as illustrated by the right side of Figure ..
Significant research has gone into each of these individual design dimensions and has resulted
in a wide range of products and technologies. Research on the Embedded Machine [] and
Kokyu [] mainly addresses the real-time dimension. The CORBA Event service [], Real-time
Publish/Subscribe [], and Distributable Threads [] provide alternative programming models
that support both one-to-many and one-to-one communication and hence address the distribution
dimension. Small footprint middleware is the main focus of e*ORB [] and UCI-Core []. TAO []
Features
Features
FIGURE .
Features, footprint, and performance.
