Mobile Agents - IME-USP

Mobile Agents:
A Key for Effective
Pervasive Computing
Roberto Speicys Cardoso & Fabio Kon
University of São Paulo - Brazil
Pervasive Computing Properties

Networking – unstable and energy
consuming

Heterogeneous

Dynamic and Complex

Seamless integration between computers
and the real world
2
Why Mobile Agents?




Reduce network load
Flexible
Can perform tasks on behalf of the
administrator or application
Use of mobile agents can lead to advantages
in terms of performance, flexibility and
escalability.
3
Proposed solution

We propose a mobile-agent enabled
solution for:

System adaptation based on service migration

Automatic updates for software components

QoS negotiation
4
System Architecture

Manages 3 aspects

Adaptation

Software Evolution

Quality of Service Negotiation
5
6
7
8
9
Service Adaptation






Dynamic environment
Services must be very responsive
Computer power available
Services implemented as mobile agents
Monitor the environment to detect changes
on resource availability
Migrates or spans a clone to a more suitable
host
10
Component Updates





High number of devices
New component versions must be pushed to the
devices
Components must be able to pull other
components on which they depend
Mobile agents encapsulate component data and
commands for installation
They can traverse a pre-defined path of nodes,
stoping on each node
11
Quality of Service Negotiation





Multimedia will be a central part of the user
interface
Benefits of service migration cannot be lost
QoS negotiation may require long periods of
connected communication
Mobile agents can avoid connected
communication
They can roam the nodes to negotiate QoS
contracts
12
Current Status

Service capable of adapt itself is under
development, using Aglets

A framework developed in our research group
will be used for activity monitoring

A prototype will be ready by the end of the
first semester of 2003
13
Related Work

one.world (University of Washington)


Pervasive Computing (IBM)


Doesn’t provide mechanisms for component
updates or means QoS negotiation
Doesn’t provide mechanisms for resource
reservation or service migration for adaptation
Gaia (University of Illinois)

Approach complementary
14
Conclusion

Mobile agents are a powerful tool for
pervasive computing environments

They are a key factor to increase the
flexibility, scalability and performance of such
systems

The ongoing research on pervasive
computing is neglecting these aspects
15