Software Architecture Taxonomies - Behaviour: Components & Connectors

Software ArchitectureS

ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Software ArchitecturePart II. Software Architecture TaxonomiesLesson 8: Components and connectors

Jose Emilio Labra Gayo

2014

ENEnglish

University of Oviedo


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Styles based on behaviour

Runtime behaviourComponents and connectors


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Main classification

Data flowJob managementInteractionRepositoryInvocationDistributed systemsAdaptable systems


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Data flowBatchPipes & Filters

Pipes & Filters with uniform interface


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Batch

Independent programs are executed sequentially

Data is passed from one program to the next

NoteBatch style can be considered the grandfather of software architectural styles

Stage

Write port

Stage

Stage

ConnectorStage

Read port


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Batch

Elements: Independent executable programs

ConstraintsOutput of one program is linked to input of

the nextA program usually waits for the previous one

to finish its execution


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Batch

AdvantagesLow coupling between componentsRe-configurabilityDebugging

It is possible to debug each input independently


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Batch

ChallengesIt does not offer interactive interfaceRequires external interventionNo support for concurrency

Low throughputHigh latency


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Pipes & Filters

Data flows through pipes and is processed by filters

Filter

Filter

Filter

Pipe

Filter

Filter Filter

Write port

Read port


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Pipes & Filters

ElementsFilter: component that transforms data

Filters can be executed concurrently

Pipe: Takes output data from one filter to the input of another filterProperties: buffer size, data format, interaction

protocol


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Pipes & Filters

ConstraintsPipes connect outputs from one filter to inputs

of other filtersFilters must agree on the exchange format they

admit


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Pipes & FiltersAdvantages

Better understanding of global systemTotal behavior = sum of each filter behavior

Reusability: Filters can be recombined

Evolution and extensibility: It is possible to create/add new filtersIt is possible to substitute old filters by new ones

IndependenceIndependent verification of each filter

PerformanceIt enables concurrent execution of filters


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Pipes & Filters

ChallengesPossible delays in case of long pipesIt may be difficult to pass complex data

structuresNon interactivity

A filter can not interact with its environment


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Pipes & Filters

ApplicationsUnix

who | wc -lJava

Clases java.io (PipedReader, PipedWriter)

Yahoo Pipes


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo Pipes & Filters - uniform

interfaceVariant of Pipes & Filters where filters have

the same interfaceElements

The same as in Pipes & FiltersConstraints

Filters must have a uniform interface


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi


interfaceAdvantages:

Independent development of filtersRe-configurabilityFacilitates system understanding

Challenges:Performance can be affected if data have to

be converted to the uniform interfaceMarshalling


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi


interfaceExamples:

Unix operating systemPrograms with a text input (stdin) and 2 text

outputs (stdout y stderr)Web architecture: REST


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Work organizationMaster-Slave


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Master-Slave

Master divides work in sub-tasksAssigns each sub-task to different nodesThe computational result is obtained as the

combination of the slaves results results

Slave 1 Slave 2

Master

Slave N. . .

Problem

task 1 task 2 task N

Solution

result Nresult 2result 1


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Master-Slave

ElementsMaster: Coordinates executionSlave: does a task and returns the result

ConstraintsSlave nodes are only in charge of the

computationControl is done by the Master node


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Master-Slave

AdvantagesParallel computationFault tolerance

ChallengesDifficult to coordinate work between slavesDependency on Master nodeDependency on physical configuration


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Master-Slave

Applications: Process control systemsEmbedded systemsFault tolerant systemsSearch systems


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Interactive systemsMVC: Model - view - controllerMVC variants

PAC: Presentation - Abstraction - Control


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MVC

MVC: Model - View - ControllerProposed by Trygve Reenskaug (end of 70's)Solution for GUIController separates model from the view"Mental model" offered through views


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MVCElements

Model: represents business logic and stateView: Offers state representation to the userController: Coordinates interaction, views and

model

MentalModel

Controller

Model

View 1

View 2

User


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MVC

ConstraintsController processes user events

Creates/removes viewsHandles interaction

Views only show valuesModels are independent of controllers/views


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MVC

AdvantagesIt supports multiple views of the same modelViews synchronizationSeparation of concerns

Interaction (controller), state (model)It is easy to create new views and controllersEasy to modify look & feelCreation of generic frameworks


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MVC

ChallengesIncreases complexity of GUI developmentCoupling between controllers and viewsControllers/Views should depend on a model

interface Some difficulties for GUI tools


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MVC

ApplicationsLots of web frameworks follow MVC

Ruby on Rails, Spring MVC, Play, etc.Some variants

Push: controllers send orders to viewsRoR

Pull: controllers receive orders from viewsPlay


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MVC variants

PACModel-View-PresenterModel View ViewModel...


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

PAC

PAC: Presentation-Abstraction-ControlHierarchy of agentsEach agent contains 3 components

PresentationAbstraction Control

PAC Agent


PAC Agent


PAC Agent


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

PAC

ElementsAgents with

Presentation: visualization aspectsAbstraction: data model of an agentControl: connects presentation and abstraction

components and enables communication between agents

Hierarchical relationship between agents


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

PAC

ConstraintsEach agent is in charge of some functionalityThere is no direct communication between

abstraction and presentation in each agentCommunication through the control

component


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

PAC

AdvantagesSeparation of concerns

Identifies functionalitiesSupport for changes and extensions

It is possible to modify an agent without affecting others

MultitaskAgents can reside in different threads,

processes or machines


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

PAC

ChallengesComplexity of the system

Too many agents can generate a complex structure which can be difficult tom maintain

Complexity of control componentsControl components handle communicationQuality of control components is important for

the whole quality of the systemPerformance

Communication overload between agents


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

PAC

ApplicationsNetwork monitoring systemsMobile robotsDrupal is based on PAC

RelationshipsThis patterns is related with MVC

MVC has no agent hierarchyThis pattern was re-discovered as

Hierarchical MVC


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

RepositoryShared dataBlackboardRule based


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Shared data

Independent components access the same stateApplications based on centralized data

repositories

Component Component

SharedData

Component...


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Shared data

ElementsShared data

Database or centralized repositoryComponents

Processors that interact with shared data

Component Component

SharedData

Component...


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Shared data

ConstraintsComponents interact with the global stateComponents don't communicate between each

otherOnly through shared state

Shared repository handles data stability and consistency


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Shared data

AdvantagesIndependent components

They don't need to be aware of the existence of other components

Data consistencyCentralized global stateUnique Backup of all the system state


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Shared data

ChallengesUnique point of failure

A failure in the central repository can affect the whole system

Distributing the central data can be difficultPossible bottleneck

Inefficient communicationProblems for scalability

Synchronization to access shared memory


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Shared data

ApplicationsLots of systems use this approach

Some variantsThis style is also known as:

Shared Memory, Repository, Shared data, etc.BlackboardRule based systems


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Blackboard

Complex problems which are difficult to solveKnowledge sources solve parts of the problemEach knowledge source aggregates partial

solutions to the blackboard

KnowledgeSource

KnowledgeSource

Blackboard

KnowledgeSource ...

Control


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Blackboard

ElementsBlackboard: Central data repositoryKnowledge source: solves part of the problem

and aggregates partial resultsControl: Manages tasks and checks the work

state

Fuente conocimien

to

Fuenteconocimiento

Blackboard

FuenteConocimient

o ...Control


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Blackboard

ConstraintsProblem can be divided in partsEach knowledge source solves a part of the

problemBlackboard contains partial solutions that

are improving


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Blackboard

AdvantagesCan be used for open problems

ExperimentingFacilitates strategy changes

Knowledge source can be reusedSupport for fault tolerance


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Blackboard

ChallengesIt may be difficult to debugNo warranty that the right solution will be

foundIt may be difficult to establish a control

strategyLow performance

Sometimes it may need to review the incorrect hypothesis

High development costImplementation of parallelism

It is necessary to synchronize blackboard access


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Blackboard

ApplicationsSome speech recognition systems

HEARSAY-IIPattern recognitionWeather forecastsGamesAnalysis of molecular structure

Crystalis


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Rule based systems

Variant of shared memoryShared memory = Knowledge base

Contains rules and facts

InferenceEngine

Knowledge base

Rules + factsUser

Interface


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Rule based systems

Elements:Knowledge base: Rules and facts about some

domainUser interface: Queries/modifies knowledge

baseInference engine: Answers queries from data

and knowledge base

InferenceEngine

Knowledge base

Rules + factsUser

Interface


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Rule based systems

Constraints:Domain knowledge captured in knowledge

baseLimit imperative access to knowledge base

It is based on rules like:IF antecedents THEN consequent

Limits expressiveness with regards to imperative languages


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Rule based systems

AdvantagesDeclarative solution

It may be easy to modify the knowledge baseSpecially tailored to be modified by domain

experts

Separation of concernsAlgorithmDomain knowledge

Reusability


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Rule based systems

ChallengesRules Debugging PerformanceRules creation and maintenance

IntrospectionAutomatic rule learningRuntime update of rules


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Rule based systems

ApplicationsExpert systemProduction systems

Rules libraries in JavaJRules, Drools, JESS

Declarative, rule based languagesProlog (logic programming)


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

InvokationCall-returnClient-ServerEvent based architectures

Publish-SubscribeActor models


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Call-return

A component calls another component and waits for the answer

Component A Component B

call

return


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Call-return

ElementsComponent that does the callComponent that sends the answer

ConstraintsSynchronous communication:

The caller waits for the answer

Componente A

Componente B

call

return


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Call-return

AdvantagesEasy to implement

ChallengesProblems for concurrent computation

If component is blocked waiting for the answerIt can be using unneeded resources

Distributed environmentsLittle utilization of computational capabilities


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Client-Server

Variant of layers2 layers physically separated (2-tier)

Functionality is divided in several serversClients connect to services

Interface query/answer

Network

query

answer

clientserver


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Client-Server

ElementsServer: offers services through a

query/answer protocolClient: does queries and process answersNetwork protocol: communication

management between clients and serversNetwork

query

answer

clientserver


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Client-Server

ConstraintsClients communicate with servers

Not the other wayClients are independent from other clientsServers don't have knowledge about clientsNetwork protocol establishes some

communication warranties Network

query

answer

clientserver


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Client-Server

AdvantagesServers can be distributedSeparation of functionality between

clients/serversIndependent developmentScalability

General functionality available for all clientsAlthough not all the servers need to offer all the

functionality


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Client-Server

ChallengesEach server can be a single point of failiure

Server attacksUnpredictable performance

Dependency on the system and the networkProblems when servers belong to other

organizationsHow can quality of service be warranted?


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Client-Server

VariantsStatelessReplicated serverWith cache


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Network

Client-Server stateless

ConstraintServer does not store information about

clientsSame query implies same answer

query

answer

clientserver


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Client-Server stateless

AdvantagesScalability

ChallengesApplication state management

Client must remember requestsSome strategies can be established to handle

information between requests


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Network

Replicated server

ConstraintSeveral servers offer the same service

Offer the client the appearance that there is only one server

query

answerclient

server

server

server

Abstract

Server


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Replicated server

AdvantagesBetter answer timesLess latencyFault tolerance

ChallengesConsistency management between

replicated serversSynchronization


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Client-server with cache

Cache = mediator between client/serverStores copies of previous answers to the

serverWhen a query is received it return the cached

answer without asking the original server

Networkquery

answerclient server

Cache


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo


Elements:Intermediate cache nodes

ConstraintsSome queries are directly answered by the

cache nodeCache node has a policy for answer

managementExpiration time


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo


Advantages:Less network overload

Lots of repeated requests can be stored in the cache

Less answer timeCached answers arrive earlier


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo


ChallengesComplexity of configurationExpiration policyNot appropriate for certain domains

If high fidelity of answers is neededExample: real time systems


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Event based

EDA (Event-Driven-Architecture)

EventProducer

EventProcessor

EventConsumer


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Event based

Elements:Event:

Something that has happened (≠ request)Event producer

Event generator (sensors, systems, ...)Event consumer

DB, applications, scorecards, ...Event processor

Transmission channelFilters and transforms events


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Event based

Constraints:Asynchronous communication

Producers generate events at any momentConsumers can be notified of events at any

momentRelationship one-to-many

An event can be sent to several consumers


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Event based

AdvantagesDecoupling: event producer does not depend

on the consumer, nor vice versa.Timelessness: events are published without

any need to wait for the termination of any cycle

Asynchronous: In order to publish an event there is no need to finish any process


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Event based

ChallengesNon sequential executionDifficult to debug


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Event based

ApplicationsEvent processing networksEvent-Stream-Processing (ESP)Complex-event-processing

VariantsPublish-subscribeActor models


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Publish-subscribe

Components that subscribe to a channel to receive message from other components

Component

Event BusSubscribePort

PublishPort

Component

Component Component

Component


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Publish-subscribe

Elements:Component:

Component that subscribes to a channelPublication port

It is registered to publish messagesSubscription port

It is registered to receive some kind of messages

Event bus (message channel): Transmits messages to subscribers


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Publish-subscribe

Constraints:Separation between subscription/publication

portA component may have both ports

Non-direct communicationAsynchronous communication in generalComponents delegate communication

responsibility to the channel


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Publish-subscribe

AdvantagesCommunication quality

Improves performanceDebugging

Low coupling between componentsConsumers do not depend on publishers...nor vice versa...


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Publish-subscribe

ChallengesIt adds a new indirection level

Direct communication may be more efficient in some domains

Complex implementationIt may require COTS


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Actor modelsUsed for concurrent computation

Actors instead of objectsThere is no shared state between actorsAsynchronous message passing

Theoretical developments since 1973 (Carl Hewitt)


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Actor models

ElementsActor: computational entity with state

It communicates with other actors sending messages

It process messages one by oneMessagesAddresses: Identify actors (mailing address)


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Actor models

ConstraintsAn actor can only:

Send messages to other actorsMessages are immutable

Create new actorsModify how it will process next message

Actors are decoupledReceiver does not depend on sender


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Actor models

Constraints (2)Local addresses

An actor can only send messages to known addresses

Because they were given to it or because he created them

Parallelism:All actions are in parallelNo shared global stateMessages can arrive in any order


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Actor models

ChallengesMessage sending

How to handle arriving messagesActor Coordination

Non-consistent systems by definitionTransparency and scalability

Internal vs external addressesNon-local actor models

Web ServicesMulti-agent systems


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Actor models

ImplementationsErlang (programming language)Akka (library)

ApplicationsReactive systemsExamples: Ericsson, Facebook, twitter


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Distributed and network systemsBrokerPeer-to-peerMapReduce

Network


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Broker

Intermediate node that manages communication between a client and a server

Client ServerBrokerstubskeleto

n


n

bridge


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Broker

ElementsBroker

It manages communicationClient: Sends requests

Client Proxy: stubServer: Returns answers

Server Proxy: skeletonBridge: Can connect brokers


n

bridge


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Broker

AdvantagesSeparation of concerns

Delegates low level communication aspects to the broker

Separate maintenanceReusability

Servers are independent from clientsPortability

Broker = low level aspectsInteroperability

Using bridges


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Broker

ChallengesPerformance

It adds an indirection layerDirect communication can be quicker

It can generate a high coupling between components

Broker = single point of failure


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Broker

ApplicationsCORBA and distributed systemsAndroid uses a variation of Broker pattern


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Peer-to-Peer

Equal and autonomous nodes (peers) that communicate between them.

Network

Peer

Peer

Peer

Peer

Peer

Peer


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Peer-to-Peer

ElementsComputational nodes: peers

They contain their own state and control thread

Network protocolConstraints

There is no main nodeAll peers are equal


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Peer-to-Peer

AdvantagesDecentralized information and controlFault tolerance

There is no single point of failureA failure in one peer does not compromise the

whole system


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Peer-to-Peer

ChallengesKeeping the state of the system

Complexity of the protocolBandwidth Limitations

Network and protocol latencySecurity

Detect malicious peers


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Peer-to-Peer

Popular applicationsNapster, BitTorrent, Gnutella, ...This architecture style is not only to share

filese-Commerce (B2B)Collaborative systemsSensor networks...

VariantsSuper-peers


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce

Proposed by GooglePublished in 2004Internal implementation by Google

Goal: big amounts of dataLots of computational nodesFault toleranceWrite-once, read-many

Style composed of:Master-slave Batch Big

Data

Map Reduce

Result


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce

ElementsMaster node: Controls execution

Node tableIt manages replicated file system

Slave nodesExecute mappers, reducersContain replicated data blocks

Big Data

Map Reduce

Result


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - Scheme

INspired by functional programming2 components: mapper and reducer

Data are divided for their processingEach data is associated with a key

Transforms [(key1,value1)] to [(key2,value2)]

c1Input: [(key1,value1)]

v1

c1 v1

c1 v1

Output: [(key2,value2)]

c2 v2

c2 v2

c2 v2

c2 v2

MapReduce


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - Mapper

Type: (clave1, valor1) [(clave2,valor2)]

c1 vi1

c2 vi2

c3 vi3

k1 v1

k2 v2

k1 v3

k3 v4

k1 v5

k1 v6

k3 v7

mapper

mapper

mapper


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - Merge and sort

System merges and sorts intermediate results according to the keys

k1 v1

k2 v2

k1 v3

k3 v4

k1 v5

k1 v6

k3 v7

k1 v1 v3 v5 v6

k2 v2

k3 v4 v7

Mergeand sort


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - Reducer

Type: (clave2, [valor2]) (clave2,valor2)

k1 v1 v3 v5 v6

k2 v2

k3 v4 v7 reducer

reducer

reducer vf1

vf2

vf3

k1

k2

k3


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - general scheme

c1 vi1

c1 vi1

c1 vi1mappe

r

reducer

reducer

reducer vf1

vf2

vf3

k1

k2

k3

mapper

mapper

k1 v1

k2 v2

k1 v3

k3 v4

k1 v5

k1 v6

k3 v7

k1 v1 v3 v5 v6

k2 v2

k3 v4 v7

Mergeand sort

MapReduce


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - count words

d1 a b

d2 a c a

d3 a c

4

1

2

a

b

c

a 1

b 1

a 1

c 1

a 1

a 1

c 1mapper

mapper

mapper

reducer

reducer

reducera 1 1 1 1

b 1

c 1 1

Mezcla y

ordena

MapReduce

// return each work with 1mapper(d,ps) { for each p in ps: emit (p, 1)}

// sum the list of numbers of each wordreducer(p,ns) { sum = 0 for each n in ns { sum += n; } emit (p, sum)}


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo MapReduce - execution

environmentExecution environment is in charge of:

Planning: Each job is divided in tasksPlacement of data/code

Each node contains its data locallySynchronization:

reduce tasks must wait map phaseError and failure handling

High tolerance to computational nodes failures


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - File system

Google developed a distributed file system - GFS

Hadoop created HDFSFiles are divided in chunks2 node types:

Namenode (master), datanodes (data servers)Datanodes store different chunks

Block replicationNamenode contains metadata

Where is each chunkDirect communication between clients and

datanodes


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - File system

Namenode

file1: (B1 – N1 N2, B2 – N1 N2 N3)file2: (B3 – N2 N3, B4 – N1 N2)file3: (B5 – N1 N3)

B1 B1

B4

B4

B5

B5

B3 B3

Client1

Cliente

N1 N2 N3

DatanodesSlaves

B2

B2 B2

Data published by Google (2007)

200+ clusters

Lots of clusters 1000+ machines

Pools with thousands of clients

4+ PB

HW/SW fault tolerance

Master


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - Inverted index

Given a number of documents, obtain a list of words associating each word with the document where it appearsOrder documents according to number of

word appearance

Documento 1En un lugar de la Mancha , de cuyo nombre no quiero acordarme no ha mucho

tiempo que vivía un hidalgo de los de lanza en astillero,

adarga antigua, rocín flaco y galgo corredor. Una olla de

algo más…

InvertedIndex

lugar doc16, doc21, doc23, doc45Mancha doc22, doc2, doc4, doc9, doc11Quijote doc22, doc1, doc2, doc7

. . .


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - Inverted index

Basic scheme of a search engine

P1P1

P1Cacheddocuments

Web crawlin

g

WebIndexing

Index

word1 doc1, doc23, doc4,…word2 doc54,doc23word3 doc1,doc7,d1oc9,doc5...word4 doc7,doc9,doc10…

Query

Search


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo MapReduce - Inverted index

d1 a b

d2 a c a

d3 a c

d2

d1

d2

a

b

c

a d1

b d1

a d2

c d2

a d2

a d3

c d3mapper

mapper

mapper

reducer

reducer

reducera d1 d2 d2 d3

b d1

c d2 d3

Mezcla y

ordena

MapReduce

// return each word with its documentmapper(d,ps) { for each p in ps: emit (p, d)}

// sort list of documentsreducer(p,ds) { ds1 = sort(ds) emit (p, ds1)}

d1 d3

d3


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce - Popular songs

Given the listening logs of a music server, obtain the number of times that each song is listenedInspired by last.fm

2/3/2010 9:41 Ana C12/3/2010 9:42 Dani C22/3/2010 9:44 Ana C22/3/2010 10:01 Luis C12/3/2010 10:10 Ana C32/3/2010 10:15 Ana C22/3/2010 10:20 Dani C22/3/2010 10:21 Luis C42/3/2010 10:24 Luis C22/3/2010 10:26 Luis C42/3/2010 10:27 Ana C4

Analyzer

C1 2 listeners, 2 timesC2 3 listeners, 5 timesC3 2 listeners, 2 timesC4 2 listeners, 3 times


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce: popular songs9:41 A C1

MapReduce

reducer

reducer

reducer

reducer

C1 2

C2 3

C3 1

C4 2

9:42 D C2

9:44 A C2

10:01 L C1

10:10 A C3

10:15 A C2

10:20 D C2

10:21 L C4

10:24 L C2

10:26 L C4

2

5

3

10:27 A C4

mapper

mapper

mapper

mapper

C1 A

C2 D

C2 A

C1 L

C3 A

C2 A

C2 D

C4 L

C2 L

C4 L

C4 A

C1 A C

C2 D

C3 A

C4 L L

Mezcla y

ordenaA A D L

A


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo MapReduce: Friends in

commonFind the list of friends in common

Inspired by Facebook

Ana Dani Juan LuisDani Ana Juan Luis MarJuan Ana Dani Luis MarLuis Ana Dani Juan MarMar Dani Juan Luis

If Ana visits Juan's page, system should show Dani, Luis

Dani

JuanLuis

Mar

Ana

A D J L

D A J L M

J A D L M

L A D J M

M D J L


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo MapReduce: Friends in

commonA D J L

D A J L M

J A D L M

reducer

reducer

reducer

MapReduce

L A D J M

M D J L

A D D J L

A J D J L

A L D J L

A D A J L M

D J

D L

D M

mapper

mapper

mapper

mapper

D M D J L

J M D J Lmapper

A J A D L M

D J

J L

J M

L M D J L

A L A D J M

D L

J L

L M

A J L M

A J L M

A J L M

A D L M

A D L M

A D L M

A D J M

A D J M

A D J M

reducer

reducer

reducer

reducer

reducer

reducer

A D J L

A J D L

A L D J

D J A L M

D L A J M

D M J L

J L A D M

J M D L

L M D J

A D D J L A J L M

A J D J L A D L M

A L D J L A D J M

D J A J L M A D L M

D L A J L M A D J M

D M A J L M D J L

J L A D L M A D J M

J M A D L M D J L

L M A D J M D J L

Mezcla y

ordena


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce: Similarity

From the listening logs, find the number of songs in common between 2 usersInspired by Amazon (similar books),

Facebook (possible friends), etc.It is usual to do several mapReduce jobs


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce2

MapReduce: Similarity9:41 A C1

9:42 D C2

9:44 A C2

10:01 L C1

10:10 A C3

10:15 A C2

10:20 D C2

10:21 L C4

10:24 L C2

10:26 L C4

10:27 A C4

A C C1, 1

A D C2, 2

A L C2, 1

A L

mapper

mapperC4, 2

A C C1, 1

A D C2, 2

A L C2 1 C4 2

Mezcla y

ordena

reducer

reducer

reducer

A C 1

A D 2

A L 3

C1 A C

C2 D

C3 A

C4 L L

A A D L

A

MapReduce1


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce: Applications

Lots of applications:Google, 2007, 20petabytes/day, around 100,000

mapreduce jobs/dayPageRank algorithm can be implemented as

MapReduceSuccess stories:

Automatic translation, similarity, sorting, ...Other companies: last.fm, facebook, Yahoo!,

twitter, etc.


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

MapReduce: Implementations

Google (internal)Hadoop (open source)CloudMapReduce (based on Amazon

services) Aster Data (SQL)Greenplum (SQL)Disco (Python/Erlang) Holumbus (Haskell)


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo MapReduce:

Libraries/languagesHive (Hadoop): query language inspired by

SQLPig (Hadoop): specific language that can

define data flowsCascading: API that can specify distributed

data flowsFlume Java (Google) Dryad (Microsoft)


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Adaptable SystemsPluginsMicrokernelReflectionInterpreters and DSLMobile codeCode on demand


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

PluginPlugin

Plugins

It allows to extend the system using plugins that add new functionality

Runtimeengine

Base system

Plugin


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Plugins

ElementsBase system:

System that allows pluginsPlugins: Components that can be

added/removed dynamicallyRuntime engine:

Starts, localizes, initializes, executes, and stops plugins

PluginPlugin

Runtimeengine

Base system

Plugin


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Plugins

ConstraintsRuntime engine manages pluginsSystem can add/remove pluginsSome plugins can depend on other plugins

The plugin must declare dependencies and the exported API

PluginPlugin

Runtimeengine

Base system

Plugin


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Plugins

AdvantagesExtensibility

Application can get new functionalities in some ways that were not foreseen by the original developers

PersonalizationApplication can have a small kernel that is

extended on demandPlugin

Plugin

Runtimeengine

Base system

Plugin


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Plugins

ChallengesConsistency

Plugins must be added to the system in a sound way

PerformanceDelay searching/configuring plugins

SecurityPlugins made by third parties can compromise

securityPlugin management and dependencies


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Plugins

ExamplesEclipse Firefox

TechnologiesComponent systems: OSGi


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Microkernel

Identify minimal functionality in a microkernel

Extra functionality is added using internal servers

External server handles communication with other systems

Microkernel

AdapterServidor internoServidor

internoInternalserver

External

serverClient

System


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Microkernel

ElementsMicrokernel: Minimal functionalityInternal server: Extra functionalityExternal server: Offers external APIClient: External application

Adapter: Component that establish communication with external server

Microkernel

AdapterServidor internoServidor

internoInternalserver

External

serverClient

System


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Microkernel

Constraints:Microkernel implements only minimal

functionalityThe rest of the functionality is implemented

using internal serversCommunication with clients by external

servers


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Microkernel

AdvantagesPortability

It is only needed to port the kernelFlexibility and extensibility

Adding new functionality with new internal servers

Security and reliabilityCritical parts of the system are encapsulatedErrors in external parts don't affect the

microkernel


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Microkernel

ChallengesPerformance

A monolithic can be more efficientDesign complexity

Identify components in the microkernelIt may be difficult to separate parts to internal

servers

Unique point of failureIf microkernel fails, the whole system may fail


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Microkernel

ApplicationsOperating systemsGamesEditors


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Reflection

It allows to change the structure and behavior of an application dynamically

Systems that can modify themselves

Base level

Metalevel

Meta-objectProtocol


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Reflection

ElementsBase level: Implements application logicMetalevel: Aspects that can be modifiedMetaobject protocol: Interface that can

modify the metalevel

Base level

Metalevel

Meta-objectProtocol


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Reflection

ConstraintsBase level uses metalevel aspects for its

behaviorAt runtime, it is possible to modify the

metalevel using the metaobject protocol

Base level

Metalevel

Meta-objectProtocol


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Reflection

AdvantagesFlexibility

The system can adapt to changing conditionsIt is not necessary to modify source code or to

stop the execution to change the running system

Base level

Metalevel

Meta-objectProtocol


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Reflection

ChallengesImplementation

Not all the languages enable meta-programming

More difficult to combine with static type systems

PerformanceIt may be necessary to do some optimizations

to limit reflectionSecurity:

Consistency maintenanceBase level

Metalevel

Meta-objectProtocol


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Reflection

ApplicationsMost dynamic languages support reflection

Scheme, CLOS, Ruby, Python, ....Intelligent systemsSelf-modifiable code

Base level

Metalevel

Meta-objectProtocol


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Interpreters and DSLs

Include a domain specific language (DSL) that is interpreted by the system

Context

InterpreterDSL

program

Application

User


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo


ElementsInterpreter: Module that executes the programProgram: Written in the DSL

DSL can be designed so the end user can write programs

Context: Environment where the program is executed

Contexto

Intérprete

Programa

en DSL

Aplicación

Usuario


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo


ConstraintsInterpreter runs the program interacting

with the contextIt is necessary to define a DSL

Syntax (grammar, parsing,...)Semantics (behavior)


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo


AdvantagesFlexibility

Adapt application behavior to user needsUsability

End users can write their own programsAdaptability

Easy to adapt to unforeseen situations


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo


ChallengesDesign of the DSLComplexity of implementation

InterpreterSeparation of context/interpreter

PerformancePossible programs may be not optimal

SecurityHandle wrong programs


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo


Variants:Embedded DSLs


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Embedded DSLs

Embedded DSLsDomain specific languages that are

embedded in general purpose host languagesThis technique is popular in soma languages

like Haskell, Ruby, Scala, etc.Advantages:

Reuse of host language syntaxAccess to libraries and IDEs of host language

ChallengesSeparation between DSL and host language


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile code

Code that is transferred from one machine to anotherSystem A sends a program to be run by system

BSystem B must contain an interpreter for the

language in which the program is written

Network

SystemA

SystemB

Interpreter

program


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile code

ElementsInterpreter: Runs the codeProgram: Program that is transferredNetwork: Transfers the program

Network

SystemA

SystemB

Interpreter

program


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile code

ConstraintsThe program must be run in the receiver

systemThe network protocol transfers the program

Network

SystemA

SystemB

Interpreter

program


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile code

AdvantagesFlexibility and adaptability to new

environmentsParallelism

ChallengesComplexity of implementationSecurity

Network

SystemA

SystemB

Interpreter

program


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile code

VariantsCode on demandRemote evaluationMobile Agents


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Network

Code on demand

Code is downloaded and run by the clientCombination between mobile code and

client-serverExample:

ECMAScript

Client

Program

ServerQuery

InterpreterAnswer


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Code on demand

ElementsClientServerCode that is transferred from server to client

ConstraintsCode resides or is generated by the serverIt is transferred to the client when it asks for

itIt is run by the client

Client must have an interpreter for the corresponding language


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Code on demand

AdvantagesImproves user experienceExtensibility: Application can add new

functionalities that were not foreseenNo need to install or download a whole

applicationAlways Beta

Adaptability to client environment


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Code on demand

ChallengesSecurityCoherence

It may be difficult to ensure an homogeneous behavior in different types of clientsClient can even decide not to run the program

Reminder: Responsive design


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Code on demand

Applications:RIA (Rich Internet Applications)

HTML5 standardizes a lot of APIsImproves coherence between clients

VariantsAJAX

Initially: Asynchronous Javascript and XMLThe program that is running at the client side

sends asynchronous requests to the server without stopping its running


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Remote evaluation

System A sends program to system B to be run and obtain its results

SystemA

SystemB

Answer(Result)

Interpreter

Program

Query

Network


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Remote evaluation

ElementsSender: Does the query including the

programReceiver: Runs the program and returns the

resultsConstraints

Receiver runs the programIt must contain some interpreter of the

program language or the program could be in machine code

Network protocol transfers program and results


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Remote evaluation

AdvantagesExploits capabilities of third parties

Computational capabilities, memory, resources, etc.

ChallengesSecurity

Untrusted codeVirus = variant of this style

Configuration


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Remote evaluation

Example: Volunteer computation

SETI@HOMEIt was the basis for the BOINC system

Berkeley Open Infrastructure for Network Computing

Other projects: Folding@HOME, Predictor@Home, AQUA@HOME, etc.


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile agents

Code and data can move from one machine to another to be runThe process takes its state from machine to

machineCode can move autonomously

Systema B

Program Interpreter

System A

Program Interpreter


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile agents

ElementsMobile agent: Program that travels and is

run from one machine or another autonomously

System: Execution environment where the mobile agents are run

Network protocol: transfers state between agents

Systema B

Program Interpreter

System A

Program Interpreter


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile agents

ConstraintsSystems host and run mobile agentsMobile agents can decide to change its

running from one system to anotherThey can communicate with other agents

Systema B

Program Interpreter

System A

Program Interpreter


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile agents

AdvantagesIt can reduce network traffic

Code blocks that are run are transmittedImplicit parallelismFault tolerance to network failuresAgents can be conceptually simple

Agent = independent unit of executionIt is possible to create mobile agent systems

Adaptability to environtment changesReactive and learning systems


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile agents

ChallengesComplexity of configurationSecurity

Malicious or incorrect code

Systema B

Program Interpreter

System A

Program Interpreter


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

Mobile agents

ApplicationsInformation retrieval

Web crawlersPeer-to-peer systemsTelecommunicationsRemote control and monitoring

Systems:JADE (Java Agent DEvelopment framework)IBM Aglets


ch

ool

of

Com

pu

ter

Scie

nce

Un

ivers

ity

of

Ovi

edo

End of presentation

Technology

Software Architecture Taxonomies - Behaviour: Components & Connectors