Image Processing Reference
In-Depth Information
This high-level view should be reflected by the data model and services of the network protocol.
For the node application programmer, a so-called shared-variable model is particularly convenient.
It is a good match for the data flow centric nature of this high-level view and quite naturally allows
separation of the node application from its communication bindings. On every change of a specially
designated variable (possibly holding the present value of an output data point), the nodes' system
sotwarewillautomaticallyinitiatethenecessarymessageexchangetopropagatetheupdatedvalue
to the appropriate receivers. his style of communication is also referred to as “implicit messaging,”
as opposed to “explicit messaging.”
Communication services roughly fall into three categories: process data exchange, management,
andengineering/diagnostics.Notallofthemarenecessarilyavailablewithinaparticularsystem.
Process data exchange
(process communication) comprises aspects of transferring data related
to I/O and processing functions. Endpoints may have command/event semantics (at-most-once)
or state/value semantics (at-least-once). Command invocations may be associated with priorities.
Value transfer may be initiated by the data sink (cyclic polling) or by the data source (likewise in a
time-triggered manner, or in response to a change of value).
Communication relationships for process data exchange are static and can be established at system
setup time. he addresses of the source and sink(s), respectively, thus will be predefined in the node.
hisisalsoknownasidentiiedcommunication.
As an alternative to the client-server paradigm, data sources can use content-based communica-
tion, using tags that describe the type of content of their transmissions instead of recipient addresses.
In this case, sources do not know their associated sinks (producer-consumer paradigm). his mode
of communiation is also identified, since the tags are necessarily static.
Management services
support management functions and those processing functions related to
building services equipment maintenance. Communication services for monitoring of data points
are of particular importance.
Communication relationships are less static than for process communication. Communication
is more often initiated on-demand, the communication endpoints chosen ad hoc. Publisher-
subscriber-based services which allow data sinks to subscribe with the data source, specifying the
criteria of interest (e.g., minimum or maximum limit), are well suited to this purpose. With these
services, the list of sinks served by a source is managed dynamically, either by the source itself or a
separate publisher entity.
Engineering and diagnostics services
support setup and maintenance of the distributed automation
system itself. Services must be provided to support the binding process as well as device application
downloads (configuration) and setting of application parameters (parametrization). Identification
services are useful to designate nodes as targets for subsequent engineering actions by way of manual
interaction (e.g., pushing a button on the device that is to receive a program download). Example
services related to diagnostics are heartbeats (and their monitoring), which can be used to detect
device or cable failures, and counters for failed transmissions.
In addition to manual configuration of bindings, system concepts may include support for devices
to provide self-binding capabilities. Usually, the system integrator is responsible that the semantics
of bound communication endpoints are compatible. Automatic binding schemes may use standard-
ized identifiers for particular matching combinations of communication endpoints to replace this
knowledge. Such an approach necessarily reduces flexibility as it requires a more stringent high-level
application model.
Ad hoc operation, as it is typical for engineering and diagnostics services, means that communica-
tion relationships are not predeined. hus, these as well as management services are best supported
by services which allow the devices (or communication endpoints) present on the network to be
discovered automatically. In addition, networks should be able to provide descriptive information
about themselves by providing communication services that enable the attributes to be associated
with devices and communication endpoints, e.g., type identifiers (possibly including the associated
