Embedded Software

1
Embedded Software

• Original Paper by Edward A. Lee
• Eal_at_eecs.berkeley.edu
• Presentation and Review by Paresh .P. Bafna
• bafna_at_usc.edu

2
What is Embedded Software?

• Software
• Realization of mathematical functions.
• Like a black box, takes input gives output.
• Not much concerned about non-functional aspects.
• Embedded software
• It executes on machines, which are not computers
  e.g.
• Its principle role interact with physical world.
• Its works at low level, as its build by experts
  of that application domain e.g.
• Its designed for specific application and so its
  not scalable e.g.
• It has to consider the non-functional aspects.

3
Non Functional Properties
4
Just software on small computers?

• Timeliness Concurrency
• Computations take time and primary goal is gain
  control over time.
• Plus doing jobs simultaneously i.e. concurrently.
• Various ways to implement concurrency are
• Interleaving processes.
• CPU bound and I/O bound processes.
• Various ways to implement timeliness are
• Priority.
• Statistical speedups e.g. cache schema, branch
  prediction etc
• Liveness
• Turing view of computation Vs Embedded view of
  computation.
• Programs never terminate.

5
Just software on small computers? cont...

• Interfaces
• Component technology exposes interface and hides
  complexity.
• Embedded software should inherit this interface
  with dynamic properties e.g.
• One form of interface is a procedure. Buts its
  finite computation.
• Similarly, OO design uses procedures along with
  data. Provide interface.
• Lee argues that Embedded software is more like
  processes than procedures.
• However, processes get weak, due to porcess1
  process2 ! process3.
• To allow dynamic composition, type systems along
  with temporal prop are used.
• Heterogeneity
• The system mixes computational styles and
  implementation technologies i.e.
• Reactivity
• System reacts to various stimuli at the speed of
  the environment. e.g.

6
Limitations of present SE methods.

• Complex Embedded system would benefit if
  components were composing other components. Hide
  details and expose interfaces in well defined
  manner.
• Procedures and Object Orientation
• Procedures are terminating. Objects are passive.
• Real world is active and more like processes,
  with proper semantics rooted in physical world.
• Embedded software requires is composition of
  processes and concurrency among processes.
• Hardware design
• Distinction among hardware and software, from the
  perspective of computation
• An application with concurrency and temporal
  content should be thought using hardware
  abstractions.
• An application with no temporal content should be
  thought using software abstraction.
• Embedded software requires system needs of both
  hardware and software.

7
Limitations of present SE methods. Cont

• Real time operating system
• Embedded software requiring real-time throughput
  also require low latency.
• Most general purpose processors have dedicated
  resources e.g.
• Embedded system are resource constrained and have
  to use scheduling techniques to allocate
  resources at run-time.
• Key problem in scheduling is the composition of
  the components.
• A chronic priority based scheduling might lead to
  e.g. priority inversion.
• Real-time Object oriented models
• These are real-time practices extended for
  distributed components e.g. real time CORBA.
• Scheduling is done by associating priority with
  event handling and other parameters (importance
  etc..).

8
Actor Oriented Design

• What do we need?
• Approach which constructs complex applications by
  composing components.
• Emphasizing concurrency, communication
  abstractions along with time constraints.
• What do we have?
• Actor oriented design, where components are actor
  with ports and parameters.
• Ports for interaction among actors with proper
  semantics (not call and return).
• Actor oriented design described in two ways
• Abstract syntax system is decomposed into
  interconnection components without defining its
  meaning.
• Concrete syntax various abstract syntax along
  with concrete define how the components fit
  together.

9
Actor oriented design cont

• Semantics of Actor oriented design
• Component is a process, and connector represents
  connection between process.
• Component may be a state and connector be
  transition between states.
• Semantics may be viewed as architectural patterns
  called models of computations.
• Key challenges for embedded software to choose
  models of computations
• Support concurrency.
• Network systems require communication and
  bandwidth.
• Applications may be interconnection of modules
  and modules written in c, java etc
• Large applications may mix various models of
  computations.

10
Examples of models of computation

• Data Flow Model
• Actors are atomic components triggered by
  availability of incoming data.
• Connection within actors represents data flow
  from producer to consumer.
• Time Triggered
• Connection is represented by events driven by
  some clock.
• Components communicate synchronously with other
  components.
• Synchronous/Reactive Model
• Connections represents input and output data
  values aligned with global clock.
• Its not necessary that every signal must have a
  value at every clock tick.
• Discrete Event Model
• Connections represented by set of events placed
  on time line, having values and time stamp.
• Specially designed for hardware or communication
  systems

11
Examples of models of computation cont

• Process Network
• Connection represent sequence of data or values
  of token.
• Components that map input data to output using
  asynchronous buffered mechanism.
• Rendezvous Model
• Components communicate in synchronous atomic
  instantaneous action.
• Processes exchange data simultaneously, in single
  step.
• Publish and subscribe Model
• Connections represented by named event streams.
• Consumer component registers and interest in the
  stream.
• Producer component produces and notifies the
  consumer component..
• Continuous Time Model
• Differential equations coupled together to find
  solutions (time constraints).
• Connection represents continuous time signals.

12
Examples of models of computation cont

• Finite state Machine
• Different from all the above model, it is
  strictly sequential.
• Component in the model is a state or mode. Only
  one state active at a time.
• Connection between states represents transition.
• It is used for control logic in embedded system
  and in-depth analysis.
• FSM can be hierarchically combined with other
  concurrent model to get hybrid model called
  chart.

13
Case Study Ptolemy II

• A project at Berkeley emphasizes on heterogeneous
  combination of models of computation.
• Composition of model is formed via the notion of
  domain polymorphism.
• What is domain polymorphism? Its a property of
  components.
• The components are used in several domain, their
  interface is different for different domain.
• Application is a set of composed actor, connected
  together, assigned a domain.
• Govern interaction between components and flow of
  control.
• To get hierarchical mixture of domains, the
  domain must implement executable interface. E.g.
  three phases, initialization, iteration and
  termination.

14
Component Interface

• To make the combination of models of computation
  more systematic, Lee argues that type system
  concept should be used.
• Type system constrains
• What a component can say about its interface.
• How compatibility is ensured when components are
  composed.
• In order to provide dynamic properties of an
  interface, system level type system is
  introduced.

15
Conclusion

• Embedded system requires a different view of
  computation. As the software engages the physical
  world.
• The system has to consider time and other
  non-functional properties.
• Models of computation with stronger properties
  are specialized.
• This specialization limits the capability, which
  can be overcome by hierarchically combining
  heterogeneous models of computations.
• System level types capture key features of
  components, and provides a robust and
  understandable composition technology.

16

• End of Presentation