Nikolay Stoimenov, Lothar Thiele ETH Zurich

Nikolay Stoimenov, Lothar Thiele ETH Zurich

Determining a Schedulability Region using
Modular Performance Analysis and Real-Time
Interfaces in WSN
Nikolay Stoimenov, Lothar Thiele
ETH Zurich
1
Outline
• Motivation for worst-case analysis of WSN
• Modular Performance Analysis / Real-Time
Calculus
• Real-Time Interfaces and constraints
propagation
• Case study: distributed MPEG decoder
2
Motivation
• WSN – diverse distributed embedded
systems running applications with diverse
requirements
3
Motivation 2: Applications
• Some of them are safety-critical
applications with strict real-time
constraints on performance
– Health care systems
– Industrial monitoring
– Intrusion detection
– Fire detection
4
Motivation 3: Analysis Questions
Memory requirements
at each node?
Throughput requirements for the network?
End-to-end delays?
5
Modular Performance Analysis / Real-Time Calculus
• Modular Performance Analysis: framework
for worst-case performance analysis of
distributed embedded systems
– Based on Real-Time Calculus [Thiele, et al.]
and Network Calculus [Le Boudec, Thiran]
– Provides worst-case guarantees on memory,
timing, and throughput constraints
– Analytical method that can be used in designspace exploration
6
Modular Performance Analysis / Real-Time Calculus 2
Resources availability
WS Node
Processing semantics
and functionality
Output stream
WS
Node
t
Input stream
Abstract Model
Δ
β
α
α’
RTC
7
Interface-Based Design
• Typical analysis question: “Does a
particular system meet its real-time
requirements?”
• Would like to ask design questions: “What
are the min/max throughput rates
supported by a system?” , “What are the
min resource requirements for a node?”
given certain memory and time constraints
8
Real-Time Interfaces
• Real-Time Interfaces: an interface theory based
on Real-Time Calculus and Assume/Guarantee
Interfaces [Henzinger, et al.]
• Can answer the question: “Given a partially
designed system and certain real-time
constraints, what are the assumptions towards
the environment such that the system still works
(the constraints are satisfied)?”
• Provides a general framework for propagating
constraints in a system
9
Case Study
10
Case Study: Analysis Questions
Do buffers overflow ?
Do buffers overflow or underflow ?
11
Case Study: Design Questions
What are feasible input rates ?
What are feasible
processor speeds ?
What are feasible
scheduling policies ?
12
Case Study: EDF Scheduling in PE2
Constraints
What are the possible min/max
deadlines for EDF scheduling ?
13
Case Study: Results
14
Conclusion
• Framework for design and analysis of
distributed embedded systems based on
Real-Time Calculus and Real-Time
Interfaces
– Worst-case analysis of timing, memory, and
throughput constraints
– Allows to answer design questions through
propagation of constraints
• Implementation: Real-Time Calculus
MATLAB Toolbox: www.mpa.ethz.ch
15
Questions and Answers
• Thank you
• Nikolay Stoimenov, Lothar Thiele:
[email protected]
16