Dynamic, Adaptive, Scalable and Robust Application Framework

1
Dynamic, Adaptive, Scalable and Robust
Application Framework

• Keita Fujii

2
Problem

• Problems with current application framework
• Static
• The function of an application is fixed (Static)
• Some applications are configurable or expandable
  by using module/plugin.
• Not robust
• If some part of the application is corrupted, it
  cant recover by itself
• For example, EXCEL may not start if graph drawing
  component is corrupted, even user doesnt want to
  draw a graph
• Not available
• You need to install an application to every
  machines in order to use it anywhere

3

• Not adaptable
• You need to choose a proper package (version)
  according to your machine spec, your language,
  function etc
• You need WORD for PDA, WindowsCE, Windows
• You need to use different package for different
  spec/language/function
• Not scalable
• Current application distribution mechanism is
  based on centralized concept, so not scalable
• Server/Client type application is not scalable
  because

4
Goal

• Future application framework should be
• Dynamic/Adaptable
• Function, requirement and quality of an
  application should be dynamic to adapt users
  demand, location, preference, device spec, etc
• Robust
• Application should work correctly or should
  recover even if some part of it may corrupt
• Available
• User should be able to use his/her favorite
  application anywhere, anytime
• Scalable
• Application must be available not matter how high
  the demand is

5
How? Combine 4 approaches

• Service Composition
• Compose service dynamically from components to
  satisfy user request, preference, location,
  device spec etc
• Generate a new or extend an existing service
  structure to satisfy variety of users demand
• First, lets focus on on the fly type
  mechanism, as opposed to pre-grouping ideas.
  
• Transparent Component Network Layer
• (I dont know if this is a proper term or not)
• Provide access to a component regardless of its
  location and status
• Is this like discovery mechanism (like
  Michaels and Enotmoto kuns?)
• Maintain consistence between distributed objects
• Detect corrupted/malicious component and replace
  by substitute one
• This can be combined with the first item
  (access to a component). Thinking about a
  distributed manner to do this is very
  interesting.
• Monitor transaction and recover automatically if
  transaction is failed

6

• Adaptive and Robust Component Distribution
• Allocate (distribute) service according to users
  demand
• using, say, energy?
• Increase or decrease number of components
  according to users demand
• Using energy?
• (Basically these are the features of Bionet)
• Global Lookup Service
• Scalable and efficient information distribution /
  advertisement framework over dynamic network
• something like Michael/Enomoto discovery
  mechanism?
• Provide up-to-date information
• Short response time
• Complex query/response (more than keyword)

7
Overview of entire architecture
Service Composition
Global Lookup Service
Mobile Agent Middleware
Transparent Component Network Layer
Adaptive Component Distribution
P2P Network
Network Layer (TCP/IP, Bluetooth)
8
Some Research Ideas

• Service Composition
• Use Service Graph (Structure of Service) as DNA
• I dont like an idea of service graph.
• Component may have Service Graph
• Service is constructed based on Service Graph
• Service Graph may change according to users
  evaluation
• Service Graph may expand, shrink, combined when
  its owner replicates, reproduces
• Component having good Service DNA would survive
  and distribute widely
• Questions
• How to describe Service Graph?
• How to construct a real composite service from
  Service Graph in distributed manner?
• How to determine that the Service Graph is good
  (frequently used) or not?
• How can Service Graph description language
  support for dynamic change?
• Whats a good scheme to change Service Graph for
  evolution?

9

• Dynamic Generation of Service Graph
• can you somehow integrate this with on the
  fly type service composition (Itao-sans
  version)?
• By Interface Matching
• Need to design an interface definition language
  that is suitable for dynamic interface matching
• Loose Interface Definition by XML Schema
• By Semantics Matching
• Need to define a semantics description language
• 5W1H?
• RDF?
• DARPA MIL?
• Task Oriented Language?
• By User
• System has to provide to user a function to
  combine any of components as he wants
• System has to reflect users evaluation about
  combination of components to determine whether it
  should be a new composite service or not
• By expanding existing Service graph

10

• Transparent Component Network
• Effective and guaranteed message routing for
  active mobile agent over ad-hoc network -gt
  There may be a lot of research already done on
  this topic (in ad hoc nets, etc.)
• Scalable and effective synchronization of
  distributed objects
• same here, as well. There may be a lot of
  research done. Are you thinking of a new
  approach?
• Adaptive and Robust Component Distribution
• Comparison of evolution mechanisms using
  weight-value and decision tree
• Global Lookup Service
• Information Distribution by Mobile Agents in
  Circular Path