Real-Time Systems. Design Principles for Distributed Embedded Applications. Herman Kopetz. Second Edition (811374), страница 34
Текст из файла (страница 34)
As anyvariable, it consists of the two parts, the static variable name and the dynamic variablevalue. The MSD contains the static variable name (let us assume the variable is namedTemperature-11) and the position where the dynamic variable value is placed in anarriving bit stream of the message. The meta-level specification explains the meaning ofTemperature-11 (see also the examples in Sect. 2.2.4).4.7Component IntegrationA component is a self-contained validated unit that can be used as a building blockin the construction of larger systems.
In order to enable a straightforward composition of a component into a cluster of components, the following four principles ofcomposability should be observed.4.7.1Principles of Composability1. Independent Development of Components: The architecture must support theprecise specification of the linking interface (LIF) of a component in the domains4.7 Component Integration103of value and time. This is a necessary prerequisite for the independent developmentof components on one side and the reuse of existing components that is based solelyon their LIF specification on the other side.
While the operational specification ofthe value domain of interacting messages is state-of-the-art in embedded systemdesign, the temporal properties of these messages are often ill defined. Many of theexisting architectures and specification technologies do not deal with the temporaldomain with the appropriate care. Note that the transport specification and theoperational LIF specification are independent of the context of use of an opencomponent, while the meta-level LIF specification of an open component dependson the context of use.
Interoperability of open components is thus not the same asinterworking of open component, since the latter assumes the compatibility of themeta-level specifications.2. Stability of Prior Services: The stability of prior services principle states thatthe services of a component that have been validated in isolation (i.e., prior tothe integration of the component into the larger system) remain intact after theintegration (see the example in Sect. 4.4.1).3.
Non-Interfering Interactions: If there exist two disjoint subgroups of cooperating components that share a common communication infrastructure, then thecommunication activities within one subgroup must not interfere with thecommunication activities within the other subgroup. If this principle is notsatisfied, then the integration within one component-subgroup will depend onthe proper behavior of the other (functionally unrelated) component-subgroups.These global interferences compromise the composability of the architecture.Example: In a communication system where a single communication channel is shared byall components on a first-come first-serve basis, a critical instant is defined as an instant,when all senders start sending messages simultaneously. Let us assume that in such asystem ten components are to be integrated into a cluster. A given communication system iscapable to handle the critical instant if eight components are active.
As soon as the ninthand tenth component are integrated, sporadic timing failures are observed.4. Preservation of the Component Abstraction in the Case of Failures: In acomposable architecture, the introduced abstraction of a component mustremain intact, even if a component becomes faulty. It must be possible todiagnose and replace a faulty component without any knowledge about thecomponent internals. This requires a certain amount of redundancy for errordetection within the architecture.
This principle constrains the implementationof a component, because it restricts the implicit sharing of resources amongcomponents. If a shared resource fails, more than one component can be affectedby the failure.Example: In order to detect a faulty component that acts like a babbling idiot, thecommunication system must contain information about the permitted temporal behaviorof every component. If a component fails in the temporal domain, the communicationsystem cuts off the component that violates its temporal specification, thus maintaining thetimely communication service among the correct components.1044.7.24 Real-Time ModelIntegration ViewpointsIn order to bring an understandable structure into a large system, it makes sense toview – as seen from an integration viewpoint – a (original) cluster of components asa single gateway component.
The integration viewpoint establishes a new clusterthat consists of the respective gateway components of the original clusters. Viewedfrom an original cluster, the external (local) interface of the gateway become theLIFs of the new cluster, while, viewed from the new cluster, the LIF of the gatewayto the original cluster is the local interface of the new cluster (see Sect. 4.5.2). Thegateways to the new cluster make only those information items available to the newcluster that are of relevance for the operation of the new cluster.Example: Figure 4.1 depicts a cluster of components that form the control system withinan automobile. The vehicle-to-vehicle gateway component (the right lower component inFig.
4.1) establishes a wireless link to other vehicles. In this example, we distinguish thefollowing two levels of integration: (1) the integration of components into the clusterdepicted in Fig. 4.1 and (2) the integration of a car into a dynamic system of cars that isachieved via the car-to-car (C2C) gateway component. If we look at the integration ofcomponents within the cluster of Fig. 4.1, then the communication network interface (CNI)of the C2C gateway component is the cluster LIF. From the C2C communication viewpoint, the cluster LIF is the (unspecified) local interface of the C2C gateway component(see also the last paragraph of Sect.
4.5.2).The hierarchical composition of components and clusters that leads to distinctintegration levels is an important special case of the integration of components.Multi-levelness is an important organizing principle in large systems. At the lowestintegration level primitive components (i.e., components that are considered to beatomic units and are not composed any further) are integrated to form a cluster. Onedistinct component of this cluster is a gateway component that forms, together withdistinct gateway components of other clusters a higher level cluster.
This process ofintegration can be continued recursively to construct a hierarchical system withdistinct levels of integration (see also Sect. 14.2.2 on the recursive integration ofcomponents in the time-triggered architecture).Example: In the GENESYS [Obm09, p.
44] architecture, three integration levels areintroduced. At the lowest level, the chip level, the components are IP cores of an MPSoCthat interact by a network on chip. At the next higher level, chips are integrated to form adevice. At the third integration level devices are integrated to form closed or open systems.A closed system is a system in which the subsystems that form the systems are known apriori. In an open system subsystems (or devices) join and leave the system dynamically,leading to a system-of-systems.4.7.3System of SystemsThere are two reasons for the rising interest in systems-of-systems: (1) the realization of new functionality and (2) the control of the complexity growth of large4.7 Component Integration105systems caused by their continuous evolution.
The available technology (e.g., theInternet) makes it possible to interconnect independently developed systems (legacysystems) to form new system-of-systems (SoS). The integration of different legacysystems into an SoS promises more efficient economic processes and improvedservices.The continuous adaptations and modifications that are necessary to keep theservices of a large system relevant in a dynamic business environment brings abouta growing complexity that is hard to manage in a monolithic context [Leh85]. Onepromising technique to attack this complexity problem is to break a single largemonolithic system up into a set of nearly autonomous constituent systems that areconnected by well-defined message interfaces.
As long as the relied-upon propertiesat these message interfaces meet the user intentions, the internal structure of theconstituent systems can be modified without any adverse effect on the global systemlevel services. Any modification of the relied-upon properties at the message interfaces is carefully coordinated by a coordination-entity that monitors and coordinatesthe system evolution. The SoS technology is thus introduced to get a handle on thecomplexity growth, caused by the necessary evolution of large systems, by introducing structure and applying the simplification principles of abstraction, separation ofconcerns, and observability of subsystem interactions (see Sect.
2.5).The distinction between a system of sub-systems and a system of systems is thusbased on the degree of autonomy of the constituent systems [Mai98]. We call amonolithic system made out of subsystems that are designed according to a masterplan and are in the sphere of control of a single development organization a systemof sub-system. If the systems that cooperate are in the sphere of control of differentdevelopment organization we speak of a system of (constituent) systems.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
