Real-Time Systems. Design Principles for Distributed Embedded Applications. Herman Kopetz. Second Edition (811374), страница 63
Текст из файла (страница 63)
Around half of all devicefailures are caused by thermal stress.The most effective way to reduce the power consumption of CMOS devices isthe scaling of the device parameters, i.e., making the transistors smaller.The scaling factor a from one micro-electronic generation to the next is normally0.7 such that the area of a scaled version of a design is reduced by a factor of 2,the powerpffiffiffi requirement is reduced by a factor of 2, thepffiffiffispeed is increased by afactor 2 and the energy performance increases by 2 2.The positioning of all subsystems onto a single die (SoC) leads to a significantreduction of the distances between the transistors in the diverse subsystems (and212lllllllll8 Power and Energy Awarenessin consequence to a reduction of the capacities of the signal lines) which resultsin major energy savings.Device scaling cannot go on indefinitely because there are limits that have theircause in the discrete structure of matter and quantum mechanical effects, such aselectron tunneling.If the frequency for the operation of the device is reduced, the voltage can bereduced as well without disturbing the functionality of the device, resulting insubstantial energy savings.Real-time and execution time are two different time bases that are not closelyrelated.A technology-agnostic design methodology makes it possible to move functionality from software to hardware with a substantial gain in energy efficiency.Pollack’s rule states that micro-architectural advances from one generation tothe next increase the performance of sequential processors only with the squareroot of the increase in power or silicon area.Spatial locality of computational subsystems reduces the effective capacitanceand thus increases energy efficiency.The most important contribution of software to energy efficiency is a reductionof the number and types of statements that must be executed to achieve thedesired results.The actual amount of energy that can be retrieved from a battery depends on thedischarge pattern.
If the requested battery power is highly irregular, the batteryefficiency is reduced and the actual energy that can be drawn out of a battery canbe less than half of the nominal energy.Energy harvesting refers to techniques that transform ambient energy (e.g.,photovoltaic, temperature gradients, electric fields, mechanical motion, vibrations, wind, etc.) to electric energy that can be used to drive low-power electronic devices, such as a wearable computer or a node in a sensor network.Bibliographic NotesThe tutorial survey by Benini and Micheli [Ben00] gives an excellent overview ofsystem-level design methods that lead to energy-efficient electronic systems.
Pedram and Nazarian [Ped06] provide models to investigate the thermal effects insubmicron VLSI circuits. The limits of device scaling are the topic of [Fra01].Pollack’s rule and the future of SoC architectures are the topic of [Bor07]. Issues ofenergy awareness in systems-on-chip are discussed in [Pol07].Review Questions and Problems8.1 Explain the difference between power and energy. Give an example where areduction of the power leads to an increase of energy needed to complete acomputation.Review Questions and Problems2138.2 How many Joule are contained in a calorie or in a kWh?8.3 Calculate the dynamic energy of a program execution if the program contains1,000,000 instructions, the supply voltage is 1 V, and the effective capacitanceof an average instruction is 1 nF.8.4 What is static energy? How does static energy change with temperature?8.5 How much energy is required for access to the scratchpad memory, the on-chipmemory, and the off-chip memory in the reference architecture introduced inthis chapter?8.6 A sensor node executes 100,000 instructions per second (supply voltage is 1 Vand effective capacitance on an instruction is 1 nF) and sends every second amessage with a length of 32 bytes to its neighbor node, which is 10 m away.The voltage of the transmitter is 3 V.
How much power is needed to drive thesensor node? How many hours will the sensor node operate if the power supplycontains two AAA batteries?8.7 A processor has two operating modes, a time-performance optimized modecharacterized by a voltage of 2 V and a frequency of 500 MHz and an energyoptimized mode characterized by a voltage of 1 V and a frequency of 200 MHz.The effective capacity of an instruction is 1 nF. What is the power requirementin each of the two modes?8.8 A Lithium-Ion laptop battery weighs 380 g. How long will a battery-load last ifthe laptop has a power demand of 10 W?Chapter 9Real-Time Operating SystemsOverview In a component-based distributed real-time system we distinguish twolevels of system administration, the coordination of the message-based communication and resource allocation among the components and the establishment, coordination, and control of the concurrent tasks within each one of the components.
Thefocus of this chapter is on the operating system and middleware functions within acomponent.In case the software core image is not permanently residing in a component (e.g.,in read-only memory), mechanisms must be provided for a secure boot of thecomponent software via the technology-independent interface. Control mechanismsmust be made available to reset, start, and control the execution of the componentsoftware at run-time. The software within a component will normally be organized ina set of concurrent tasks. Task management and inter-component task interactionshave to be designed carefully in order to ensure temporal predictability and determinism. The proper handling of time and time-related signals is of special importancein real-time operating systems.
The operating system must also support the programmer in establishing new message communication channels at run time and incontrolling the access to the message-based interfaces of the components. Domainspecific higher-level protocols, such as a simple request-reply protocol, that consistof a sequence of rule-based message exchanges, should be implemented in themiddleware of a component. Finally, the operating system must provide mechanismsto access the local process input/output interfaces that connect a component to thephysical plant.
Since the value domain and the time-domain of the RT-entities in thephysical plant are dense, but the representation of the values and times inside thecomputer is discrete, some inaccuracy in the representation of values and timesinside the computer system cannot be avoided. In order to reduce the effects of theserepresentation inaccuracies and to establish a consistent (but not fully faithful) modelof the physical plant inside the computer system, agreement protocols must beexecuted at the interface between the physical world and cyberspace to create aconsistent digital image of the external world inside the distributed computer system.A real-time operating system (OS) within a component must be temporallypredictable.
In contrast to operating systems for personal computers, a real-timeH. Kopetz, Real-Time Systems: Design Principles for Distributed Embedded Applications,Real-Time Systems Series, DOI 10.1007/978-1-4419-8237-7_9,# Springer Science+Business Media, LLC 20112152169 Real-Time Operating SystemsOS should be deterministic und support the implementation of fault-tolerance byactive replication. In safety-critical applications, the OS must be certified. Since thecertification of the behavior of a dynamic control structure is difficult, dynamicmechanisms should be avoided wherever possible in safety-critical systems.9.1Inter-Component CommunicationThe information exchange of a component with its environment, i.e., other components and the physical plant, is realized exclusively across the four message-basedinterfaces introduced in Sect.
4.4. It is up the generic middleware and the component’s operating system to manage the access to these four message interfaces forinter-component communication. The TII, the LIF and the TDI are discussed in thissection, while the Local Interface is discussed in the section on process input/output.9.1.1Technology Independent InterfaceIn some sense, the technology independent interface (TII) is a meta-level interfacethat brings a new component out of the core-image of the software, the job, and thegiven embodiment, the component hardware into existence. The purpose of the TIIis the configuration of the component and the control of the execution of thesoftware within a component.
The component hardware must provide a dedicatedTII port for the secure download of a new software image onto a component.Periodically, the g-state (see Sect. 4.2.3) of the component should be published atthe TII in order to be able to check the contents of the g-state by a dedicateddiagnostic component. A further TII port directly connected to the componenthardware must allow the resetting of the component hardware and the restart ofthe component software at the next reintegration point with a relevant g-state that iscontained in the reset message. The TII is also used to control the voltage andfrequency of the component hardware, provided the given hardware supportsvoltage-frequency scaling. Since malicious TII messages have the potential todestroy the correct operation of a component, the authenticity and integrity of allmessages that are sent to the TII interface must be assured.9.1.2Linking InterfaceThe linking interface (LIF) of a component is the interface where the services of thecomponent are provided during normal operation.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
