A Framework for Effective Data Transfer

1
A Framework for Effective Data Transfer
AMobe - project

• Stijn Bernaer
• Patrick De Causmaecker
• Joris Maervoet
• Greet Vanden Berghe

2
AMobe

• IWT funded project
• Stands for Application Development for Mobile
  Devices
• Executed by the I.T. Department of KaHo
  Sint-Lieven
• 1/9/2002 -- 31/8/2004
• Partners

3
Application Development for Mobile Devices

• 2 coordinators and 2 researchers
• How are mobile devices integrated in
  applications? Based on 3 case studies
• Development of FrEDT an (agent) Framework for
  Effective Data Transfer
• FrEDT is a generic agent organisation that
  manages client-server communication in a wireless
  environment
• Runs on mobile devices with limited capacity

4
Disciplines
Agent architecture
Wireless technology
FrEDT
Agent Platforms
Platforms
Programming Platforms
Operating Systems
5
Wireless communication technology
2G

• 2.5G 2G extensions for higher bitrates
• Move from phone-oriented to data-oriented
  networks
• Move from circuit-switched to packet-switched
  data
• Average rates - GSM 9.6 kbps - HSCSD 28.8 kbps
  - GPRS 40 kbps - UMTS 384 kbps

2.5G
3G
6
Agent architecture

• Five trends have dominated computer history
• Ubiquity
• Interconnection
• Intelligence
• Delegation
• Human-orientation
• How to incorporate these trends in our
  applications?

7
Agent architecture

• An agent is a computer system that is able to
  represent its owner
• An agent can find out what it needs to realise
  its design goals
• A multi-agent system consists of communicating
  agents
• Those agents will represent owners with diverse
  interests and goals. They will have to
  collaborate, coordinate and negotiate.

8
Agent architecture

• FIPA (Foundation for Intelligent and Physical
  Agents) defines standards for heterogeneous and
  interacting agents and agent-based systems

9
Agent architecture

• FIPA agents communicate through the use of FIPA
  ACL (Agent Communication Language) messages
• Ontologies decribe the structure and semantics
  of the message content. A content language
  provides the message syntax
• Interaction protocols define possible sequences
  of high-level communicative acts (sequence of
  message types)

10
Platforms for mobile devices
11
Case 1 Transfer of geographical data

• Central geographical database
• Mobile employees create Update Reports (UR)
• Employees need to transfer these URs
• Wireless synchronisation process between local
  and central Update Report DataBase (URDB)
• Employees need the most recent Update Reports on
  regions that will be explored in the near future
  (pro-active)
• Agents regulate the data traffic

12
Case 2 Supporting people with non-congenital
brain injuries

• People who have problems with coordinating and
  organising their daily life
• Complete database on the supervisors computer
  and a subset of the data on the smartphone/PDA of
  the patient
• Days schedule, addresses, instructions,
  itineraries, shopping lists
• Optionally mobile device for supervisor
• Agents act as assistants

13
Case 3 Transfer of medical reports and analyses

• Mobile employees visit companies for risk
  analysis, ergonomic analysis and medical
  inspection
• Employees use a procedure handbook and a program
  to manage analysis records on their notebook
• This book has to be updated regularly and changed
  records need to be exchanged with a central
  database
• Agents are connection-aware and run in background

14
Framework for Effective Data Transfer

• Three cases ?FrEDT
• Generic agent organisation
• Synchronisation mechanism
• Wireless client-server environment
• Agents negotiate about transactions with
  different priorities
• Goals to avoid congestion, to find a balance
  between transaction costs and transaction pending
  time

15
Why agents?

• Component-based environment
• To assist the end user
• Asynchronity delay independence
• Autonomy, continuity
• Adaptivity, proactivity
• Negotiation

16
FrEDT Ontology

• Development of an device context ontology for
  FrEDT
• An example

17
Transactions and decisions

• Decisions about transaction planning are taken
  centrally (no local filtering)
• Aware of connection rate and cost. (If necessary,
  indicated by the user.)
• Non-urgent transactions are delayed based on
  prediction.

Centraltransactionmanagement
1 Request
Device
Server
2 Agree
3 Transaction
18
Centraltransactionmanagement
1 Request
Device
Server
2 Agree
3 Transaction
19
(No Transcript)
20
Negotiation Agent

• Represents the device interests
• Gathers knowledge about connection speed and
  costs and passes this information to the Index
  Agent
• Receives application specific transaction
  requests
• Handles further request negotiation with the
  Synchro Agent
• Delegates transactions after agreement (creates
  temporary Transaction Agents)

21
Negotiation Agent

• How does the Negotiation Agent gather knowledge
  about the connection?
• Speed mainly pings (same protocol as
  transaction) to the Index Agent.
• Costs clustering if possible, else interaction
  with the user is needed
• How often is knowledge about connection updated?
• When the device logs on
• After completed transactions
• Pings at regular times

22
Negotiation Agent

• Device should know when it is worth to notify the
  connection change to the central system
• Heikki Helin watermark technique
• FrEDT proportion of current connection to last
  registered connection

23
Transaction Agent

• Created by the Negotiation Agent
• Is responsible to handle the raw transaction
• Lifetime 1 transaction
• Independent on transaction protocol
• Separates the framework from the
  application-specific database

Database Agent
24
Index Agent

• Holds a central register of logged-on devices and
  their connections
• Devices interact with this agent to keep this
  register up-to-date
• Answers questions concerning this register

25
Prediction Agent

• Generates predictions on connections and costs
• Provides the Synchro Agent with predictions to
  assist in transaction scheduling
• Periodic model (assumes periodic behaviour) with
  recency effect

26
Synchro Agent

• Receives transaction requests from interested
  servers and devices
• Schedules transactions - decisions are made based
  on following indicators
• Application-defined priority (from Negotiation
  Agent)
• Connection rate and costs (from Index Agent) and
  their predictions (from Prediction Agent) of
  sender/receiver
• Pending time

27
Conclusion

• FrEDT is suited for effective datatransfer in
  different situations
• FrEDT joins several disciplines agents, mobile
  devices and wireless communication technology
• Points of attention
• Fault-tolerance and security (provided by the
  agent platform)
• Complex ontological transaction requests

28
Demonstration

• At the DSP Valley stand