Image Processing Reference
In-Depth Information
Publishing phase
Indication
Request
Indication
Indication
6
6
5
6
Subscription phase
Confirmation Request
Indication
Response
4
1
2
3
Subscriber
Publisher
Subscriber
Subscriber
FIGURE .
Push-type publisher-subscriber model.
statically during system setup and commissioning or dynamically during runtime. As the underly-
ing communication mechanism of the publisher-subscriber model is a multicast to a defined group
of nodes, it must be ensured that a node subscribing to a given message or object joins the correct
communication group. his is therefore mainly done using client-server-type communication based
on confirmed services.
Processes with mostly event-based communication can get along very well with publisher-
subscriber- or producer-consumer-type communication systems. Depending on the fundamental
layer- communication methods, real-time requirements may be more (in the case of TDMA meth-
ods or centralized polling) or less met (with pure random access methods). he obvious advantage
is that all connected devices have direct access to the entire set of information as the broadcasting is
based on identification of messages rather than nodes. Reaction times on events can be very short due
to the absence of slow polling or token cycles. Generally, publisher-subscriber-type systems (or sub-
systems) are mostly multi-master systems because every information source (producer) must have
the possibility to access the bus. he selection of relevant communication relationships is solely based
on message filtering at the subscriber's side. Such filter tables are typically defined during the plan-
ning phase of an installation. It is no wonder that this communication model is being widely used
in fieldbus systems, as for instance WorldFIP, Foundation Fieldbus, CAN, CANopen, DeviceNet,
ControlNet, LIN, EIB, or LonWorks.
Still, there is an inherent problem. As both paradigms are message-based and therefore connec-
tionless on the application layer, they are not suited for the transmission of sensitive, non-repetitive
data such as parameter and configuration values or commands. Connectionless mechanisms can
inform the respective nodes about communication errors on layer , but not about errors on appli-
cation layer. Particularly for fieldbus systems devised for safety-critical applications (such as TTP),
additional mechanisms have been developed to enable atomic broad- and multicasts. Atomic in this
context means that the communication is successful only if all subscribers receive the distributed
information correctly. Only in this case will actions associated with the transaction be executed,
otherwise they are canceled.
20.5.6 Above the OSI Layers—Interoperability and Profiles
A key point for the acceptance of fieldbus systems was their openness. The availability of publicly
available specifications and inally standards made it possible to set up so-called multi-vendor systems
where components developed by different manufacturers work together to achieve a common
functionality. This possibility was the evidence for interoperability and still is an important argu-
ment in fieldbus marketing. he standardization of fieldbusses was originally thought to be sufficient
for interoperable systems, but reality quickly showed that it was not. Standards (interoperability
