Composing Time- and Event-driven Distributed Real-time Systems

1
Composing Time- andEvent-driven
DistributedReal-time Systems

• Gabor Madl (gabe_at_ics.uci.edu),
• Ph.D. Candidate, UC Irvine
• Advisor Nikil Dutt (dutt_at_ics.uci.edu)
• Chancellors Professor, UC Irvine

2
Challenges in NGAS
Challenges Traditional Design Compose
Functionalities Combine Analysis
Model-based Analysis

• How to safely increase functionality?
• Primary concern is safety (at least it should be)
• Secondary concern is cost (?)
• Increase functionality while constraints above
  are preserved
• How would a painter work under these conditions?

3
Separate Functionalities
Challenges Traditional Design Compose
Functionalities Combine Analysis
Model-based Analysis

• Dedicated hardware for each functionality
• Protect components from each other
• Design them independently
• Are we sure that there is no interaction between
  critical and non-critical functionalities?
• Leakage power drains power even when the car is
  idle
• Energy consumption could become a bottleneck
• How will critical functionalities perform in a
  resource-constrained environment?
• Suboptimal utilization
• More components are needed
• Limited interaction with the environment

4
Rethink Design of NGAS
Challenges Traditional Design Compose
Functionalities Combine Analysis
Model-based Analysis

• We need to use more flexible design methodologies
  than the current practice
• We need to learn to better utilize the potential
  of distributed real-time embedded (DRE) systems
• More and more sensors and actuators
• More interaction between components and their
  environment
• We need to build on the strengths of existing
  design methodologies, but also encourage
  interaction
• Cars could use information from the environment
  (i.e. weather information, GPS, other cars) to
  prepare for unforeseen circumstances, such as
  fog, freezing, accidents ahead etc.
• Non-critical functionality could be used as
  backup to increase fault tolerance

5
Compose Functionalies
Challenges Traditional Design Compose
Functionalities Combine Analysis
Model-based Analysis

• Critical functionalities
• Time-triggered systems
• Focus on control (scheduling)
• Execution times, periods, deadlines, priorities,
  etc.
• Mathematical model for analysis (scheduling
  theory)
• Simple analysis, costly implementation
• Non-critical functionalities
• Event-driven systems
• Focus on the flow of data
• Throughput, communication architecture,
  parallelization, etc.
• Complex model, hard to predict all behaviors
• Simple implementation, costly analysis

6
Need to Combine Analysis Methods
Challenges Traditional Design Compose
Functionalities Combine Analysis
Model-based Analysis

• Static analysis methods
• Often too abstract, resulting in
  conservative/inaccurate results
• Cannot capture dynamic effects
• Simulations
• Can show the presence of an error, never its
  absence
• Ad-hoc, hard to measure coverage
• Limited design space exploration
• Model checking
• State space explosion problem
• No partial results
• Time consuming and costly
• Each method has its advantage and disadvantage

7
Model-based Design Analysis
Challenges Traditional Design Compose
Functionalities Combine Analysis
Model-based Analysis

• Model-based design provides the means for the
  early exploration of design alternatives
• The design flow is driven by the DSM, a
  high-level specification that captures key
  properties
• Mappings play a key role in abstraction
• Formal models drive functional verification
• We propose the combination of simulations and
  formal methods for the evaluation of designs

8
Questions?

• Links to relevant work
• http//dre.sourceforge.net
• http//alderis.ics.uci.edu
• http//www.ics.uci.edu/gabe