SelfConfiguration of Adaptive Ubiquitous Applications

1
Self-Configuration of Adaptive Ubiquitous
Applications

• Prof. Dr. Kurt Rothermel
• Dr. Christian Becker
• Marcus Handte

2
Outline

• Motivation
• Project Status
• Configuration
• Integration
• Summary
• Outlook

3
Motivation

• Ubiquitous Computing
• Distraction-free support for tasks
• Distributed applications
• Dynamic environments (mobility)
• ? Applications need to adapt
• A common approach
• Applications as compositions of
  services/components
• Application composition needs to be configured at
  runtime
• Research goal
• Design a lightweight algorithm that
  (re-)configures applications in
  resource-restricted environments taking into
  account the resulting costs and potential quality
  gains using a simple quality metric

Universal Messenger
4
Project Overview

• Determine a valid application configuration
• Analyze the requirements of configuration
  algorithms
• Design an algorithm that configures an
  application
• Adapt an existing application configuration
• Analyze adaptation costs and develop cost metric
• Enhance previous approach to adapt configuration
  with minimal costs
• Optimize an application configuration
• Develop suitable quality metric for optimization
• Extend algorithms to consider quality metric
• Balance between adaptation cost and quality gains

5
Project Status

• Determine a valid application configuration
• Adapt an existing application configuration
• Optimize an application configuration

• ? Problem interpretation and extended
  backtracking algorithm
• M. Handte, C. Becker, K. Rothermel Peer-based
  Automatic Configuration of Pervasive
  Applications, Journal on Pervasive Computing
  and Communications, 1(4), Troubador Publishing
  LTD, 251-264, December 2005
• M. Handte, C. Becker, K. Rothermel Peer-based
  Automatic Configuration of Pervasive
  Applications, 2nd IEEE International
  Conference on Pervasive Services (ICPS) 2005,
  Santorini, Greece, 2005
• Michael Reinsch, Entwicklung und Evaluation
  eines Algorithmus zur initialen Konfiguration von
  Anwendungen in PCOM, Diplomarbeit, Universität
  Stuttgart, Fakultät Informatik, Elektrotechnik
  und Informationstechnik, 2004

? Problem interpretation and extended
backtracking algorithm

• ? Framework support for automatic adaptation of
  stateful components
• ? Greedy cost optimization on top of initial
  backtracking algorithm
• M. Handte, G. Schiele, S. Urbanski, C. Becker,
  Adaptation Support for Stateful Components in
  PCOM, Workshop on Software Architectures for
  Self-Organization Beyond Ad-Hoc Networking at
  3rd International Conference on Pervasive
  Computing (Pervasive) 2005, München, Germany,
  2005
• Andreas Störzbach, Entwicklung und Evaluation
  eines Algorithmus zur Anpassung von Anwendungen
  in PCOM, Diplomarbeit, Universität Stuttgart,
  Fakultät Informatik, Elektrotechnik und
  Informationstechnik, 2005

• ? Analysis of scenarios and design of level-based
  quality model
• T. Weis, M. Handte, M. Knoll, C. Becker
  Customizable Pervasive Applications, 4th IEEE
  International Conference on Pervasive
  Computing and Communications (PERCOM) 2006, Pisa,
  Italy, 2006
• Christian Schwegmann, Entwicklung eines
  Rahmenwerks zur Modellierung von
  Qualitätsmerkmalen für Konfigurationen in PCOM
  , Diplomarbeit, Universität Stuttgart, Fakultät
  Informatik, Elektrotechnik und
  Informationstechnik, 2005

6
System Model

• Peer-to-Peer Approach
• Devices in communication range cooperate
• No supporting infrastructure required
• PCOM Component System
• Components with contractually specified
  dependencies towards other components and local
  resources
• Container provides runtime environment for
  components and manages (potentially limited)
  local resources
• Container provides basic services (e.g.
  communication, device discovery)
• ? Applications as tree of component instances
  composed along dependencies

Input
Output
7
Automatic Configuration
File System (Input)
PDA
PPT Control (Application)
CPU
5
MEM
10
CPU
100
Imager (Display)
MEM
100
Input
Output
DISP
1
Laptop
Local PPT (Output)
CPU
5
MEM
10
Remote PPT (Output)
CPU
5
DISP
CPU
300
1
MEM
10
MEM
200
DISP
1
Display
Display

• ? Composition that satisfies structural
  requirements (no unresolved dependencies) and
  resource requirements (no overloaded resources)

Executable Configuration
8
Requirements on Automatic Configuration

• Completeness
• Frustrated users if configuration sometimes not
  successful
• Systematic search that finds a valid
  configuration eventually
• Efficiency
• Short configuration delays since users might wait
• Applicable for a broad range of environments
• Distribution
• Do not rely on powerful server for configuration
• Suitable for peer-based systems
• Resilience
• Deal with failures and mobility during
  configuration

9
Approach

• Interpretation as Distributed Constraint
  Satisfaction Problem
• Given Set of variables (distributed across
  agents) with finite domains and constraints
  between variable assignments
• Find Valid variable assignment for all variables
  
• Adapted Asynchronous Backtracking (Yokoo et al.
  97)
• Distribution use available parallelism, no
  powerful device
• Dependency-directed reconsiders only
  assignments that can resolve a conflict (conflict
  locality)
• Extension for resilience integrated

10
Mapping to CSP

• Dependencies as variables
• Component instances as domains
• Resource and structural requirements as
  constraints
• Partial solution
• Introduction of pseudo value
• Constraints
• If instance is used, dependencies must be
  resolved
• If instance is not used, pseudo value must be
  selected
• Instances on container must not overload resources

Input
Output
11
Evaluation

• Simulation of binary tree with 12 components on 4
  containers
• 1-30 components that introduce resource
  conflicts, 1 solution
• 1 conflicting resource on each container (no
  conflict locality)

12
Integration
Components

• Demonstrator Multihop Presenter
• Pervasive Presenter(Component-based Application)
• PCOM (Component System)
• BASE (Object-oriented Middleware)
• M. Handte, S. Urbanski, C. Becker, P. Reinhard,
  M. Engel, M. Smith 3PC/MarNET Pervasive
  Presenter, 4th IEEE International Conference on
  Pervasive Computing and Communications (PerCom
  2006), Pisa, Italy, 2006

PCOM Container
Semantic Service Matching

• Semantic Service Matching (Jena)
• Service Mgmt and Discovery (Aachen)
• Mobility-aware Multicast (Karlsruhe)

Contract Framework
Configuration Algorithms
Resource Management
Efficient Service Discovery
BASE Micro-broker
Invocation Broker
Registries

• Position-based Multicast (Mannheim)
• Publish/Subscribe Middleware (Magdeburg)
• Routing Emulation(Marburg)

Plugin Manager
Communication (Routing, Multicast)
Discovery Communication Plugins
...
Emulation
Network
13
Summary

• Ubiquitous Computing requires self-configuring
  applications
• Basis Basic abstractions for adaptive
  applications (PCOM)
• Goal Provide suitable algorithms for automatic
  (re-)configuration
• Approach Map configuration to CSP and extend
  existing algorithms
• Status On schedule
• Priority Programme Integration
• Synergetic and complementary research
• Emulation, Service Discovery, Multi-hop Routing,
  Multicast, Semantic Service Matching,
• Demonstrator
• Marburg, Mannheim, Magdeburg

14
Outlook

• 3rd phase introduces new focus on heterogeneity
• Device capabilities will be heterogeneous
  (resource-poor and rich)
• Device mobility will be heterogeneous (static as
  well as mobile)
• Users tend to reuse same applications in same
  environments
• Research goal for 3rd phase
• Leverage assumptions to reduce configuration
  delay
• Proposed research direction
• Dynamically distribute configuration process
  depending on resources
• Enable dynamic distribution of configuration
  algorithms
• Analyze the effects of various degrees of
  distribution
• Derive and integrate distribution heuristic
• Reuse partial application configurations in less
  dynamic scenarios
• Analyze different models for partial
  configurations to foster efficient reuse
• Enhance algorithms to enable usage of existing

15
More Resources

• Web
• www.3pc.info
• Some references
• C. Becker, M. Handte, G. Schiele, K. Rothermel
  PCOM A Component System for Pervasive
  Computing, 2nd IEEE International Conference on
  Pervasive Computing and Communication (PerCom),
  Orlando, USA , 2004
• M. Handte, C. Becker, K. Rothermel Peer-based
  Automatic Configuration of Pervasive
  Applications, Journal on Pervasive Computing and
  Communications, 1(4), Troubador Publishing LTD,
  251-264, December 2005
• T. Weis, M. Handte, M. Knoll, C. Becker
  Customizable Pervasive Applications, 4th IEEE
  International Conference on Pervasive Computing
  and Communications (PerCom), Pisa, Italy, 2006
• E-Mail
• kurt.rothermel christian.becker
  marcus.handte _at_ ipvs.uni-stuttgart.de