A%20Mobile%20Agent%20Infrastructure%20for%20QoS%20Negotiation%20of%20Adaptive%20Distributed%20Applications - PowerPoint PPT Presentation

About This Presentation
Title:

A%20Mobile%20Agent%20Infrastructure%20for%20QoS%20Negotiation%20of%20Adaptive%20Distributed%20Applications

Description:

Aglets. Adaptation Framework. DSRT. QML ... Considered time elapsed between event notification and aglet departure to new host ... – PowerPoint PPT presentation

Number of Views:112
Avg rating:3.0/5.0

less

Transcript and Presenter's Notes

Title: A%20Mobile%20Agent%20Infrastructure%20for%20QoS%20Negotiation%20of%20Adaptive%20Distributed%20Applications


1
A Mobile Agent Infrastructure forQoS Negotiation
ofAdaptive Distributed Applications
  • Roberto Speicys Cardoso Fabio Kon
  • University of São Paulo USP
  • BRAZIL

2
Outline
  • Infrastructure for QoS negotiation
  • Highly distributed environments
  • Adaptive applications
  • Based on Mobile Agents
  • Topics
  • Architecture
  • Implementation
  • Experimental results

3
Motivation
  • Multimedia applications and Ubiquitous Computing
    services require QoS
  • Current research focus is not on QoS negotiation
  • QoS negotiation is important
  • Exchange of several messages
  • High number of QoS configurations
  • Must be effective

4
Architectural Requirements
  • Fundamental
  • Scalable, flexible and complete
  • But also...
  • Light QoS monitoring process
  • QoS specification ? Application code
  • Admission control
  • Negotiation and Re-negotiation
  • Effective
  • Secure

5
Architectural Design
  • Central monitoring management node (replicable)
  • On each node (not every node)
  • Monitoring component
  • QoS enforcement system
  • QoS broker
  • QoS negotiation system
  • Mobile agent server

6
(No Transcript)
7
Implementation
  • Reuse systems and tools whenever possible
  • Aglets
  • Adaptation Framework
  • DSRT
  • QML
  • Identify technology shortcomings when applied to
    ubiquitous computing

8
Performance Monitoring
  • Requirements
  • Applications must be aware of system-level and
    application-level QoS
  • Light monitoring process
  • Framework for Dynamic Adaptation of Distributed
    Systems
  • Applications subscribe to simple or compound
    events
  • Central node contacted only when an event of
    interest happens

9
QoS Enforcement
  • DSRT
  • Processing time and memory guarantees
  • Runs on top of SunOS, IRIX, Linux and Windows
  • QoS requirements specified using DSRT API calls
  • Our architecture decouples specification and
    source code
  • Same application may work with another QoS
    enforcement system

10
QoS Definition
  • QML used to avoid DSRT specification using API
    calls
  • QDL
  • Also good
  • Too tied to QuO
  • XML
  • Easier to implement
  • Too verbose
  • Harder to understand by humans

11
QML Definition Example
  • type CPU_RT_PCPT contract
  • period decreasing numeric usec
  • peakProcessing_per_period increasing numeric
  • idealProcessing CPU_RT_PCPT contract
  • period 40
  • peakProcessing_per_period 30
  • idealProfile for mpeg_player profile
  • require idealProcessing

12
QoS Negotiation
  • QoS levels change throughout time
  • Resource sharing, component failure, peaks of
    utilization
  • Applications need efficient means for QoS
    negotiation
  • Our framework allows applications to
  • Wait for best QoS offer or timeout and choose
    among the received replies
  • Choose roaming strategy Linear x Parallel

13
Negotiation process
  • Application subscribes to events with the
    monitoring system
  • Monitoring system notifies application
  • Application requests a QoS negotiation
  • Negotiator agent reads specification, defines
    negotiation strategy and creates ProxyNegotiator
    agents
  • ProxyNegotiator travels to remote hosts to
    negotiate and send back results
  • Negotiator collects results and passes to
    application
  • Application choses the best QoS offer

14
QoS Brokering
  • One host may negotiate with many applications
    simultaneously
  • Reservation concurrency problems
  • One application may negotiate with many hosts
    simultaneously
  • Resource starvation issues
  • CORBA QoS Broker
  • Reservation requests generates tickets
  • Ticket required for reservation
  • When ticket expires, broker releases resources

15
Adaptation Support
  • Applications should use the mobile agent paradigm
  • Flexibility and adaptability benefits
  • Two adaptation techniques Migration and Cloning
  • Migration allows application to move to hosts
    with better QoS offers
  • Cloning enables applications to adapt gracefully
    to service usage bursts and valleys

16
Sample Application
  • Multimedia reflector
  • ReflectorAglet
  • Specifies QoS requirements
  • Detects processing shortages
  • Uses QoS negotiation system
  • Adapts to QoS changes

17
Tests Performed
  • Goal Evaluate scalability
  • Estimate increase in execution time when
    negotiating with multiple hosts
  • Linear and parallel roaming strategies
  • Each test performed 5 times
  • Considered time elapsed between event
    notification and aglet departure to new host

18
(No Transcript)
19
Future Work
  • Evaluate the infrastructure with other
    distributed applications
  • Develop more sophisticated QoS negotiation
    algorithms
  • Evaluate infrastructure usage (in which scenarios
    to use it)

20
Conclusions
  • We presented a mobile-agent-based infrastructure
    for QoS negotiation
  • Goal is to be an effective mechanism for QoS
    negotiation
  • First results QoS negotiation system scaled up
    well from 1 to 4 hosts
  • Evaluating up to 8 hosts
  • Mobile agents present many potential advantages
  • Concept introduced 10 years ago, but technology
    not yet widespread
  • Object orientation took over 20 years!

21
Questions?
  • Software available at
  • www.ime.usp.br/speicys(preliminary
    version)Thank you!
Write a Comment
User Comments (0)
About PowerShow.com