International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011 A Survey on A Survey on Quality of Service Quality of Service Support on Support on Middleware Middleware -Based -Based Distributed Messaging Systems Distributed Messaging Systems Used Used in Multi Agent Systems in Multi Agent Systems Jose-Luis Poza-Luján, Jose-Luis Poza-Luján, Juan-Luis Posadas-Yagüe. José-Enrique Juan-Luis Posadas-Yagüe. José-Enrique Simó-Ten Simó-Ten University Institute of Control Systems and Industrial University Institute of Control Systems and Industrial Computing (ai2) Computing (ai2) Universitat Politècnica de València (UPV) Universitat Politècnica de València (UPV) Camino de vera, s/n. 46022 Valencia (Spain) Camino de vera, s/n. 46022 Valencia (Spain) {jopolu, jposadas, {jopolu, jposadas, jsimo}@ai2.upv.es jsimo}@ai2.upv.es (A theoretical (A theoretical approach) approach)
15
Embed
A Survey on Quality of Service Support on Middleware-Based Distributed Messaging Systems Used in Multi Agent Systems
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
A Survey onA Survey onQuality of ServiceQuality of Service Support onSupport onMiddlewareMiddleware-Based-BasedDistributed Messaging Systems UsedDistributed Messaging Systems Usedin Multi Agent Systemsin Multi Agent Systems
University Institute of Control Systems and Industrial Computing (ai2)University Institute of Control Systems and Industrial Computing (ai2)Universitat Politècnica de València (UPV)Universitat Politècnica de València (UPV)Camino de vera, s/n. 46022 Valencia (Spain)Camino de vera, s/n. 46022 Valencia (Spain)
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
Multi-agent SystemMulti-agent System
Distributed systemDistributed system
MiddlewareMiddleware
Quality of ServiceQuality of Service
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
Is possible use Is possible use the QoS in a the QoS in a
distributed MASdistributed MAS
What is the What is the support offered support offered
by the most by the most middleware in middleware in
distributed MAS distributed MAS
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
AgentsAgents ObjectsObjects
MessageMessageReal-timeReal-time
MiddlewareMiddleware
And the QoS?And the QoS?
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
attributes
parameters
policies
CommunicationSystem
AgentCo
m
unicm
at i onI
s
nterface
AgentCo
m
unicm
at i onI
s
nterface
JMS
FIPA
CORBA / DDS
ContextualizationContextualization
Attributes and parameters: direct accessAttributes and parameters: direct access
Policies: managed accessPolicies: managed access
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
Connection ManagementConnection Management
Error HandingError Handing
Message FlowMessage Flow
PerformancePerformance
MetadataMetadata
Time managementTime management
3 1 1 1
2
0
0
2
2
0
3
0
0
4
0
5
0
1
0
1
4
2
1
5
QoS objectsQoS objects QoS areaQoS area
AnalyzingAnalyzing
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
Connection ManagementConnection Management
Error HandingError Handing
Message FlowMessage Flow
PerformancePerformance
MetadataMetadata
Time managementTime management
AnalyzingAnalyzing
QoS objectsQoS objects QoS areaQoS area
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
Quality of Service
Components connection
Message flow
Timemanagement
Performance
Establishment
Management
Mode
Status
Delay
ErrorsLiveliness
Reconnection
Errors
Priority
Delivery orderMessages management
Error rate
Errors Mean up time
Components management
Max hops
Persistence
Routing
Synchronization
Transaction type
Component Resource limits
CommunicationsBandwidth
Throughput
Delay
Components management
Durability
Messages management
Lifespan
Latency
Temporal filter
Deadline
Time to live
Timestamp
referred to
connectionphase
ubication
area
ubication
Cla
ssif
yin
gC
lass
ifyi
ng
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
RecommendationsRecommendations
MinimumMinimum requirements: the requirements: the real-timereal-time approach approach Message flow: priorityMessage flow: priority Time management: deadlineTime management: deadlineMaximumMaximum requirements: MAS approach the requirements: MAS approach the FIPA+JMS+CORBA+DDSFIPA+JMS+CORBA+DDS approach approach
Message flow
Time management
Messages Components
Durability
Lifespan
Time to live
Priority
Delivery order
PersistenceMax hops
Routing
Synchronization
Transaction type
Delay
Latency
Temporal filter
Deadline
Timestamp
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
ProposalsProposals
Design the system in according to the Design the system in according to the QoS QoS requirementsrequirements.. - Select the - Select the correct middlewarecorrect middleware: CORBA, JMS, …: CORBA, JMS, … - Create the infrastructure - Create the infrastructure to provideto provide the specific the specific QoS parameters: message queues, threads with QoS parameters: message queues, threads with time management, type of synchronization…time management, type of synchronization…
Depending the Depending the contextcontext is necessary a specific set is necessary a specific set of QoS parametersof QoS parameters - Real time: minimum requirements- Real time: minimum requirements … … - Agents: maximum requirements.- Agents: maximum requirements.
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
Nideadlinedeadline icomponentAgent 1)(min
NicomponentAgentN
ithroughputthroughput
1
)(
ProposalsProposals Our experienceOur experience
Using a Using a standardized QoSstandardized QoS allows to compare allows to compare distributed MAS with a quantitative values. distributed MAS with a quantitative values.
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
ProposalsProposalsBased on the standardized QoS parameters, Based on the standardized QoS parameters, define agent-oriented QoS policies: define agent-oriented QoS policies: QoAQoA (Quality (Quality of Agent)of Agent) or or QoMAS (Quality of the MAS)QoMAS (Quality of the MAS)..
Our experienceOur experience
NQoS: number of standardized QoS implemented by the agentNQoS: number of standardized QoS implemented by the agent
NAgent: number of agents working in the distributed MASNAgent: number of agents working in the distributed MAS
NQoSQoSQoSQoSQoA ,...,, 21
NAgentQoAQoAQoAQoMAS ,...,, 21
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
ProposalsProposals Our experienceOur experience
Use the QoA to determine the impact in the Use the QoA to determine the impact in the system of the agents operations system of the agents operations Allows the system to establish the Allows the system to establish the MAS work MAS work areaarea or the distributed MAS or the distributed MAS potentialpotential..
NQoSiareaworkareawork iQoSQoA1__ )(
minmax_ )(,)()( iQoSiQoSiQoS areawork
NAgentiareaworkareawork iQoAQoMAS1__ )(
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
ConclusionsConclusionsQoS is increasing his QoS is increasing his relevancerelevance in distributed in distributed systems: systems: from a static point of viewfrom a static point of view (only used to (only used to monitoring) monitoring) to a dynamic point of viewto a dynamic point of view (used to (used to manage the system)manage the system)
Most used middleware in distributed MAS, offers Most used middleware in distributed MAS, offers a different vision of the QoS. It is necessary a different vision of the QoS. It is necessary standardize the QoSstandardize the QoS parameters and policies. parameters and policies.
It is necessary It is necessary studystudy other QoS implementations other QoS implementations to define the complete set of QoS parameters.to define the complete set of QoS parameters.
Using a standardized QoS allows to Using a standardized QoS allows to comparecompare distributed MAS with a distributed MAS with a quantitativequantitative values. values.
International Symposium on Distributed Computing and Artificial Intelligence Salamanca, 6th - 8th April 2011
Thanks for your attentionThanks for your attention