L6 - Model Driven Design Flashcards
Modeling
A model represents the reality by recreating system characteristics (replication, simulation) and therefore makes verification possible.
- PRO: easy to understand, simulation of systems, evaluation of different designs, formal methods to test code against models
- CON: difficult underlying processes can lead to unwanted behaviour still possible
Design
In model-driven design, models of hardware/software/environment are necessary.
Important are:
- guaranteed performance
- fault tolerance
- verification against model.
Analysis
By conducting reachability analysis; model checking; evaluating execution times and testing the fault tolerance; we aim to understand the system.
Which type of model is discrete in time; expressed as system of ODEs and discretized with time step T?
Dynamic Models.
Finite State Machines (FSM)
Finite State Machines (FSM) are …
- discrete in states (time is implicit)
- expressed as finite set of states
- having transitions which are activated by guard expression transition
What is temporal control?
Temporal control determines instants in the real time domain when computations have to be executed.
Synchronous Reactive Systems
Synchronous Reactive Systems produce outputs synchronously to changes of inputs.
Event Triggered Control (Reactive Systems)
Event Triggered Control (Reactive Systems) produce output for each input event. Emphasis is placed on safety and determinism but execution of events can overlap.
Principles Of Composability
- Independent Development Of Component: platform independent model (PIM) complies with Linking Interface (LIF).
- Stability Of Prior Service: services validated in isolation.
- Non-Interfering Interactions: communication of subgroups should not interfere with each other.
- Preservation Of Component Abstraction: redundancy in order to detect faulty components.
Dataflow (Model Of Computation)
Execution depends on data and has to wait until data is ready. Especially Discrete Dynamics (Labview; Simulink)
Time-Triggered Execution (Model Of Computation)
Execution is planned for each instant in time. Especially real time control (Giotto; FTOS)
Component Interaction (Model Of Computation)
Composition of data and query driven execution. Especially web servers.
Process Networks (Model Of Computation)
Processes communicate through channels that message in a queue (asynchronous). Especially in distributed systems.
Rendezvous (Model Of Computation)
Processes communicate via synchronous messages (wait until both sender and receiver is ready).
Ex.: CSP, Ada
What is a Deterministic System?
In a deterministic system the worst case response time is predictable. Sufficient for RTS.