Citation
Yang Zhao. "On the Design of Concurrent, Distributed Real-Time Systems". PhD thesis, University of California, Berkeley, August, 2009.

Abstract
Achieving determinism in distributed real-time systems is challenging, due to uncertainties in execution time, communication jitter, and resource scheduling. This dissertation presents a concurrent model of computation (MoC) for distributed real-time systems called PTIDES (pronounced “tides,” for “Programming Temporally Integrated Distributed Embedded Systems”). PTIDES uses a discrete-event (DE) model as the underlying formal semantics to achieve analyzable deterministic behavior. PTIDES programs are discrete-event models constructed as networks of concurrent components, called actors, communicating via time-stamped events. These time stamps serve as the basis to define the unique order among events. Rather than using DE models for performance modeling and simulation, where time stamps are a modeling property bearing no relationship to real time during execution of the model, PTIDES uses DE model as a specification language for real-time applications. It extends DE models with the capability of relating events that interact with the physical world with physical time. Preserving DE semantics at runtime can be challenging, since the global, consistent notion of time may lead to a total ordering of execution in a distributed system, an unnecessary waste of resources. A dependency analysis framework is presented to allow out of order processing of events without compromising determinism and without requiring backtracking. The key idea is that if two events have independent affects, formally defined through causality analysis, then they can be processed in any order. As a result, if the earlier event is delayed due to communication, processing of the later event does not need to be blocked. General event triggered real-time systems with multiple shared resources are not amenable to compile-time feasibility analysis. However, when the discrete activities can come in predictable patterns, real-time scheduling theories are applicable to many PTIDES models. This dissertation studies a sufficient condition for a PTIDES model to be feasible when the inputs to a PTIDES model are sporadic, i.e. when there is a minimum interval between any two consecutive events of the same input.

Electronic downloads

• http://www.eecs.berkeley.edu/Pubs/TechRpts/2009/EECS-2009-117.html

• HTML

  Yang Zhao. <a
  href="http://chess.eecs.berkeley.edu/pubs/775.html"
  ><i>On the Design of Concurrent, Distributed
  Real-Time Systems</i></a>, PhD thesis, 
  University of California, Berkeley, August, 2009.

• Plain text

  Yang Zhao. "On the Design of Concurrent, Distributed
  Real-Time Systems". PhD thesis,  University of
  California, Berkeley, August, 2009.

• BibTeX

  @phdthesis{Zhao09_OnDesignOfConcurrentDistributedRealTimeSystems,
      author = {Yang Zhao},
      title = {On the Design of Concurrent, Distributed Real-Time
                Systems},
      school = {University of California, Berkeley},
      month = {August},
      year = {2009},
      abstract = {Achieving determinism in distributed real-time
                systems is challenging, due to uncertainties in
                execution time, communication jitter, and resource
                scheduling. This dissertation presents a
                concurrent model of computation (MoC) for
                distributed real-time systems called PTIDES
                (pronounced �tides,� for �Programming
                Temporally Integrated Distributed Embedded
                Systems�). PTIDES uses a discrete-event (DE)
                model as the underlying formal semantics to
                achieve analyzable deterministic behavior. PTIDES
                programs are discrete-event models constructed as
                networks of concurrent components, called actors,
                communicating via time-stamped events. These time
                stamps serve as the basis to define the unique
                order among events. Rather than using DE models
                for performance modeling and simulation, where
                time stamps are a modeling property bearing no
                relationship to real time during execution of the
                model, PTIDES uses DE model as a specification
                language for real-time applications. It extends DE
                models with the capability of relating events that
                interact with the physical world with physical
                time. Preserving DE semantics at runtime can be
                challenging, since the global, consistent notion
                of time may lead to a total ordering of execution
                in a distributed system, an unnecessary waste of
                resources. A dependency analysis framework is
                presented to allow out of order processing of
                events without compromising determinism and
                without requiring backtracking. The key idea is
                that if two events have independent affects,
                formally defined through causality analysis, then
                they can be processed in any order. As a result,
                if the earlier event is delayed due to
                communication, processing of the later event does
                not need to be blocked. General event triggered
                real-time systems with multiple shared resources
                are not amenable to compile-time feasibility
                analysis. However, when the discrete activities
                can come in predictable patterns, real-time
                scheduling theories are applicable to many PTIDES
                models. This dissertation studies a sufficient
                condition for a PTIDES model to be feasible when
                the inputs to a PTIDES model are sporadic, i.e.
                when there is a minimum interval between any two
                consecutive events of the same input.},
      URL = {http://chess.eecs.berkeley.edu/pubs/775.html}
  }
  

Posted by Christopher Brooks on 13 Nov 2010.
Groups: ptolemy
For additional information, see the Publications FAQ or contact webmaster at chess eecs berkeley edu.