Software Architecture - 1

May 1, 2012

All Architectural Patterns

Data Flow Architecture

Batch Sequential Pipe and Filter Process Control Repository Blackboard

Data Centered Architecture Hierarchical Architecture

Main-subroutine Master-slave Layered Virtual Machines

All Architectural Patterns

Implicit Asynchronous Communication Architecture

Interaction-Oriented Architecture Distributed Architecture
Model-View-Controller (MVC) Presentation-Abstraction-Control (PAC) Client-server Broker Service-oriented architecture (SOA) Non-buffered event-based Implicit Invocation Buffered messaged-based

Component-based Architecture

Architectural Patterns - 1

Data Flow Architecture

Batch Sequential Pipe and Filter Process Control Repository Blackboard

Data Centered Architecture Hierarchical Architecture

Main-subroutine Master-slave Layered Virtual Machines

Data-Flow Architecture

Overview

Three subcategories
Batch Sequential Pipe and Filter Process Control

A series of transformation on successive sets of data Data sets and operations are independent of each other.

Batch Sequential

Overview

Applications

A transformation subsystem or module cannot start its process until the previous module completes its computation Data flow carries a batch of data as a whole from a module to next series of transformation on successive sets of data Data sets and operations are independent of each other. Business data processing in banking and utility billing

Batch Sequential

Validate

Sort

Update

Report

Data flow

data transformation

Batch Sequential
Rejected transaction -123 U Transaction File 222 U 111 I 333 D -123 U

Validated transaction
222 U 111 I 333 D

Sorted transaction
111 I 222 U 333 D

Validate

sort

Generate Report Reports

100 ---111 ---200 ---222 ---444 ----

Update

Updated Master file

100 --200 --222 --333 --444 ---

Master file

Batch Sequential

Using Unix Shell Script

(exec) (exec) (exec) (exec) validate trans validTrans invalids sort validTrans sortedTrans update master sortedTrans generateReports master report1 report2

Batch Sequential

Applicable Domains

Benefits

Data are batched Intermediate files are sequential files Each module reads input files and writes output files Simple division on subsystems Each subsystem can be stand-alone program Implementation requires external control No interactive interface Concurrency not supported

Limitations

Pipe and Filter

Overview

A system consists of data source, filters, pipes, and data sinks Connections between components are data streams Each data stream is a first-in-first-out buffer Each filter reads data from its input stream, processes it, and writes it over a pipe for next filter to process A filter processes data once received, does not wait until the whole data is received A filter only knows its connected pipes, does not know what are at the other end of the pipe Filters are independent of each other Pipes and filters may run concurrently

Pipe and Filter

Active filter

Passive filter

Pulls in (read) data from upstream and pushes out (write) data to downstream It lets connected upstream pipes push (write) data to it and lets downstream pipes to pull (read) data from it

Pipe and Filter

data source: Data Source filter1: Filter pipe: pipe filter2: Filter data sink: Data Sink

read Filter1 W Filter2 R write

Data Source

Filter1

Filter2

Data Sink

pipe

Pipe and Filter

Hello World

Sort & Count Filter

HL

Match (H,L) Filter

LO W

Upper case Conversion Filter

orld

H 1 L 2

Pipe and Filter

Pipe and Filter

Applicable Domains

Benefits

System can be broken into a series of steps over data streams Data format in the streams is simple, stable and adaptable Significant amount can be pipelined for performance Concurrency: all filters may operate at the same time Reusability: filters are easy to plug and play No user interaction Difficult to configure Concurrency not supported

Limitations

Data-Centered Architecture

Overview A centralized data store is shared by all surrounding software components The surrounding components are normally independent each other Significant amount can be pipelined for performance Two categories:
Repository Blackboard

Repository

Overview Data store is passive Clients of the data store are active Supports interactive data processing Clients control the computation and flow of logic

Repository

Repository

Repository

Repository

Applications
Suitable for large complex information systems where many software component clients need to access it in different ways. Data transactions to drive the control flow of computation.

Benefits:
Easy to backup and restore. Easy to add new software components Reduce the overhead of transient data

Limitations:
Centralized repository is vulnerable to failure compared to distributed repository with data replication. High dependency between the structure of the data store and its agents.

Blackboard

Overview Data store is active Clients are passive Clients are also called knowledge sources, listeners, and subscribers Two partitions Blackboard: store data (hypotheses and facts) Knowledge sources: domain-specific knowledge is stored Controller: initiate the blackboard and knowledge sources and take overall supervision

Blackboard

Collaboration

Knowledge sources register with the blackboard in advance in a publish/subscribe fashion Data changes in the blackboard trigger one or more matched knowledge sources to continue processing Data changes may be caused by new deduced information or hypothesis results by knowledge sources Knowledge sources normally do not interact with each other

Blackboard

Blackboard

Blackboard example 1

Animal identification System(KBS).

The knowledge is represented as production rules.
R1: IF animal gives milk THEN animal is mammal R2: IF animal eats meat THEN animal is carnivore R3: IF animal is mammal AND animal is carnivore AND animal has tawny color AND animal has black stripes THEN animal is tiger

A set of facts
F1: F2: F3: F4: animal eats meat animal gives milk animal has black stripes animal has tawny color

Blackboard example 1

Forward reasoning
New knowledge is added to blackboard
N1: animal is carnivore (from R2 + F1) N2: animal is mammal (from R1 + F2) IF animal gives milk THEN animal is mammal N3: animal is tiger (R3 + N3 + N2 + F3 + F4)

Blackboard example 2

Travel Consulting System

Participating agents

Airline, hotel reservation, auto rental, and attraction agents Budget, available time, locations, etc Clients fill out a initial travel form The system will respond with many optional plans for client to choose

Blackboard

Clients of the system

Blackboard example 2

Travel Consulting System

Process

A client submits a request The system stores all the data in the blackboard The blackboard makes a request to the air agent Once air reservation data is returned and stored in blackboard, the change triggers hotel, auto rental, attraction agents for a travel plans under budge and time Client chooses one of the plan The system triggers the billing process

Blackboard

Applications
Suitable for open-ended and complex problems (AI) The problem spans multiple disciplines, each of them has complete different knowledge expertise Partial, or approximate solution is acceptable to the problems.

Benefits:
Easy to add new or update existing knowledge source. Concurrency: all knowledge sources can work in parallel Reusability of knowledge source agents.

Blackboard

Limitations:
Tight dependency between the blackboard and knowledge source, Difficult to make a decision when to terminate reasoning, since only partial or approximated solutions are expected Synchronization of multiple agents is an issue. Debugging and testing of the system is a challenge.

Hierarchical Architecture

Overview

Four categories
Main-subroutine Master-slave Layered Virtual Machines

The software system is decomposed into logical modules (subsystems) at different levels in the hierarchy. Modules at different levels are connected by explicit or implicit method invocations. A lower level module provides services to its adjacent upper level modules Upper level modules invoke the methods or procedures in lower level.

Main-subroutine

Overview The main-subroutine architecture has dominated the software design methodologies for a very long time. Reuse the subroutines and have individual subroutines developed independently. Using this style, a software system is decomposed into subroutines hierarchically refined according to the desired functionality of the system. Refinements are conducted vertically until the decomposed subroutine is simple enough to have its sole independent responsibility, and whose functionality may be reused and shared by multiple callers above.

Main-subroutine

Main-subroutine

Data Flow Diagram to Main-subroutine Transform flow: incoming flow feeds data in an external format, and the data is then transformed to another format, and then the outgoing flow carries the data out Transaction flow: evaluates its incoming data, and decided which path to follow among many action paths. A transform flow is mapped by a controlling module for incoming, transform and outgoing information processing. The transaction node becomes a dispatcher control module that controls all subordinate action modules

Main-subroutine

Main-subroutine

Applicable Domains

Benefits

Data are batched Intermediate files are sequential files Each module reads input files and writes output files Easy to decompose the system based on the definition of the tasks in a top-down refinement manner Globally shared data in classical main-subroutines introduces vulnerabilities. Tight coupling may cause more ripple effects of changes as compared to OO Design.

Limitations

Master-slave

Overview

A variant of the main-subroutine architecture style that supports fault tolerance and system reliability. Slaves provide replicated services to the master, and the master selects a particular result among slaves by certain selection strategy. The slaves may perform the same functional task by different algorithms and methods or totally different functionality.

Master-slave

Master-slave

Applicable Domains

Software systems where reliability is critical.

Layered Architecture

Overview The system is decomposed into a number of higher and lower layers in a hierarchy Each layer consists of a group of related classes that are encapsulated in package, in a deployed component, or as a group of subroutines in the format of method library or header file. Also, each layer has its own sole responsibility in the system. A request to layer i +1 invokes the services provided by the layer i via the interface of layer i. The response may go back to the layer i +1 if the task is completed; otherwise layer i continually invokes services from the layer i -1 below.

Layered Architecture

Layered Architecture

Layered Architecture

Layered Architecture
FTP TCP IP
Ethernet protocol FTP protocol TCP protocol IP protocol

FTP TCP IP Ethernet

Ethernet
Physical connection

Layered Architecture

Applicable Domains

Benefits

Any system that can be divided between the application specific portions and platform specific portions . Applications that have clean divisions between core services, critical services, user interface services, etc. Incremental development based on levels of abstraction. Enhanced independence of upper layer to lower layer as long as their interfaces remain unchanged. separation of the standard interface and its implementation. Component-based technology may be used to implement the layered architecture; this makes the system much easier to allow for plug-and-play of new components. Promotion of portability: each layer can be an abstract machine deployed independently. Easy to decompose the system based on the definition of the tasks in a top-down refinement manner

Layered Architecture

Limitations

Lower runtime performance since a clients request or a response to client must go through potentially several layers. There are also performance concerns on overhead on the data marshaling and buffering by each layer. Many applications cannot fit this architectural design. Breach of interlayer communication may cause deadlocks and bridging may cause tight coupling. Exceptions and error handling is an issue in the layered architecture, since faults in one layer must propagate upwards to all calling layers.

Virtual Machine

Overview A virtual machine is built up on an existing system and provides a virtual abstraction, a set of attributes, and operations. In most cases, we find that a virtual machine separates a programming language or application environment from a execution platform. Some people say that a virtual machine looks like emulation software.

Virtual Machine

Virtual Machine

Virtual Machine

Applicable Domains

Benefits

Suitable for solving a problem by simulation or translation if there is no direct solution. Sample applications include interpreters of microprogramming, XML processing, script command language execution, rulebased system execution, Smalltalk and Java interpreter typed programming language Portability and machine platform independency. Simplicity of software development. Simulation for disaster working model. Slow execution of the interpreter due to the interpreter nature. Additional overhead due to the new layer.

Limitations