Reinventing Computing for Real Time

1
Reinventing Computing for Real Time

• Edward A. Lee and Yang Zhao
• University of California, Berkeley
• Presented By
• Sarthak Datt
• 11/20/2007

2
Outline

• Introduction
• Current Issues
• Embedded Software
• Concept of Concurrency and Time
• Models of Computation
• Imperative Concurrency Models
• Declarative Concurrency Models
• Discrete-Event Runtime Framework
• Current Areas of Research
• Analysis of the Paper
• Conclusion

3
Introduction

• Study of models of computation, software
  techniques, and analytical models for distributed
  timed systems. Also about computational systems
  that are interconnected on a network.
• Given time synchronization with some known
  precision, how does this change how distributed
  applications are designed and developed?
• How can time synchronization help with realizing
  coordinated real-time events.
• The need for Reinventing Computing for Real Time

4
Current issues

• Many of the advances in computing become part of
  the problem, not part of the solution.
• Computer architecture and software have made it
  difficult or impossible to estimate or predict
  the execution time of software
• Programming languages lack time in their
  semantics, timing requirements are only
  specified indirectly
• Operating systems rely on best effort techniques
• Present Software Techniques have time and
  concurrency properties as afterthoughts.
• And many more

5
Embedded Software

• Writing embedded software in assemble code or C,
  however they fail to specify timing requirements
  or constraints.
• Embedded software has to interact with hardware
  that is specialized to the application
• Time matters. However, in the 20th century
  abstractions of computing, time is irrelevant.
• Higher reliability standard than general purpose
  software.
• Software is often written without sufficient use
  of various interlock mechanisms.

6

• The Big Question
• What would it take to achieve concurrent
• and networked embedded software that was
• absolutely positively on time, to the
• resolution and reliability of digital logic?
• Unfortunately, everything would have to
• Change

7
Concurrency and Time

• Delivering temporal semantics in software can be
  challenging.
• Time is about the ordering of events. However, in
  embedded software time also has a metric!!
• Concurrency in software is a challenging issue
  because the basic software abstraction is not
  concurrent.
• Memory of the computer represents the current
  state of the system, instructions transform that
  state.
• The state may change on its own at any time.

8
An example..

• Embedded system uses two threads, one for each
  sensor
• y getSensorData() // From thread 1
• y getSensorData() // From thread 2
• x 0.9 x 0.1 y // From thread 1
• x 0.9 x 0.1 y // From thread 2
• print x // From thread 1
• print x // From thread 2
• Possible Solution (Atomicity)
• acquireLock() // Block until acquired
• y getSensorData() // Block for data
• x 0.9 x 0.1 y // Discount old value
• print x // Display the result
• releaseLock() // Release the lock
• Does this really works??

9
Alternate Solutions( Ptolemy Project)

• Process Networks
• Model concurrency using Kahn process networks
  model of computation
• network of sequential processes
• processes do not share memory and communicate
  asynchronously
• Discrete Event
• General environment for time-oriented simulations
  of systems such as queuing systems, communication
  networks, and hardware systems
• actors communicate by sending tokens across
  connections

10
Process Network vs. Discrete Event
Process Network
Discrete Event
11
Models of Computation

• Imperative Concurrent Modelsexample TinyOS
• Declarative Concurrent Modelsexample Simulink,
  SCADE, Lab VIEW

12
nesC/TinyOS configuration
13
Discrete-Event Runtime Framework
PTIDES ( Programming Temporally Integrated
Distributed Embedded Systems
14
What needs to be done..

• Stankovic argues that the time dimension must be
  elevated to a central principle of the system.
  Time requirements and properties cannot be an
  afterthought
• Improvement on Concurrency models
• Component architecture
• Management of time-critical operations in ways
  significantly different from prevailing software
  engineering techniques
• Actor-oriented design

15
AlarmNet (Assisted-Living And Residential
Monitoring Network) - a wireless sensor network
for smart healthcare

• An architecture for smart healthcare that will
  open up new opportunities for continuous
  monitoring of assisted-living and
  independent-living residents
• integration with existing medical practices and
  technology
• miniature, wearable sensors
• assistance to the elderly and chronic patients
• Architecture is multi-tiered, with lightweight
  sensors, mobile components, and more powerful
  stationary devices. Sensors are heterogeneous,
  and all integrate into the network

16
A smart assisted-living space, instrumented with
sensors and devices .
17
System architecture, showing roles of components
from front-end to back-end, and physical
associations among them.
18
Beehive Distributed Real-Time Databases

• Global, object oriented real-time databases
• Emphasis on adding value along four dimensions
  real-time, fault tolerance, security, and Quality
  of Service
• Used when transactions have deadlines and where
  data is valid only for a period of time
• Applications include Internet services, defense
  applications, and smart spaces

19
Croquet Project

• A free software platform and a network operating
  system for developing and delivering deeply
  collaborative multi-user online applications.
• Principally been applied to three-D shared
  immersion environments on the internet, similar
  to the ones that might be used in interactive
  networked gaming.
• Features a network architecture that supports
  communication, collaboration, resource sharing,
  and synchronous computation among multiple users.

20
Real-Time Specification for Java (RTSJ)

• Set of interfaces and behavioral specifications
  that allow for real-time programming in the Java
  programming language
• Javolution
• Real-time library aiming to make Java
  applications faster and more time predictable.
• Algorithmic parallel computing support with
  concurrent contexts.
• Context programming in order to achieve true
  separation of concerns.
• Simple yet powerful configuration management for
  applications.

21
Analysis of the Paper

• Strengths
• Addresses the problem in a comprehensive manner
• Provides good examples to explain the concept and
  providing solutions
• Weakness
• Need to look outside Ptolemy Project and
• and if possible draw comparisons.

22
Conclusion

• Most of these Technologies, use conventional
  concurrency models .
• Existing methods for addressing real-time
  computation typically deal with a portion of the
  problem of constructing and executing real-time
  programs.
• Reinventing Computing for Real Time is need.
• Real Systems for the Real World, in Real Time!!
  

23
References

• Real Time and Embedded Computing Laboratory(
  http//www.cs.virginia.edu/control/ )
• http//www.rtsj.org/
• http//www.cs.virginia.edu/wsn/medical/
• PTIDES Programming Temporally Integrated
  Distributed Embedded Systems
• Yang Zhao, Edward A. Lee, Jie Liu, October 2-4,
  2006, 2006 IEEE 1588 Conference, Gaithersburg,
  MD.
• Computing for Embedded Systems
• State of the Art Lecture, Edward A. Lee, IEEE
  Instrumentation and Measurement Technology
  Conference, Budapest, Hungary, May 21-23, 2001.

24

• Thank You