Software-Architecture_Course-Notes.pdf

Software Architecture | 2
Copyright © 2017 University of Alberta.
All material in this course, unless otherwise noted, has been
developed by and is the property of the University of Alberta. The
university has attempted to ensure that all copyright has been
obtained. If you believe that something is in error or
has been omitted, please contact us.
Reproduction of this material in whole or in part is acceptable,
provided all University of Alberta logos and brand markings remain
as they appear in the original work.
Version 0.1.0

TABLE OF CONTENTS
COURSE NOTES 1
Course Overview 4
Module 1: UML Architecture Diagrams 6
Architecture Overview and Process 6
Kruchten’s 4+1 View Model 8
Logical View 9
Process View 10
Development View 11
Physical View 11
Scenarios 11
Component Diagrams 12
Package Diagrams 14
Deployment Diagram 18
Activity Diagram 22
Module 2: Architectural Styles 25
Language-Based Systems 25
Abstract Data Types and Object Oriented Design 26
Main Program and Subroutine 27
Repository-Based Systems 30
Databases 31
Layered Systems 34
Client Server n-Tier 38
Interpreter-Based Systems 43
Dataflow Systems 45
Pipes and Filters 45
Implicit Invocation Systems 47
Event Based 47
Process Control Systems 51
Feedback Loops 51
Other Variations 52
Module 3: Architecture in Practice 55
Quality Attributes 55
Analyzing and Evaluating an Architecture 62
Availability Example 64
Architecture Trade-off Analysis Method 66
Relationship to Organizational Structure 72
Product Lines and Product Families 73
Implementing Product Lines 74
Reference Architecture 77
Course Resources 81
Course Resources 81
Glossary 82

COURSE OVERVIEW
This is the third course of the Software Design and Architecture
specialization brought to you in partnership by Coursera and the
University of Alberta. The first two courses of this specialization were
Object-Oriented Design and Design Patterns. The two courses
provide a foundation for this Software Architecture course,
although you can take each course separately.
Software Architecture covers how a software system is constructed
at the highest level. Development teams will decide how a software
system will be broken down into components that work together.
This course will show you how developers use UML to document
their architectural designs and how architecture can be understood
from multiple perspectives and through different UML diagrams.
Next, you will learn about different software architectures, which
include when and how they are used, their advantages and
disadvantages, and how they are represented visually. Finally, this
course will cover how architectures are analyzed and evaluated.
Software architecture is tied to implementation languages,
deployment environments, the structure of the development team,
the needs of the business and the development cycle, opportunities
for product lines, and the intent of the software itself.

Upon completion of this course, you will be able to:
1) Use Unified Modelling Language (UML) to document
architectural design.
2) Explain how architecture can be understood from multiple
perspectives and through different UML diagrams.
3) Explain different software architectures, including:
• when and how they are used
• their advantages and disadvantages
• how to represent them visually
4) Analyze and evaluate system architectures.
5) Explain how software architecture is tied to:
• implementation languages
• deployment environments
• the structure of the development team
• the needs of the business and development
cycle
• opportunities for product lines
• the intent of the software

MODULE 1: UML ARCHITECTURE DIAGRAMS
Architecture Overview and Process
Software architecture is the fundamental design of an entire
software system. It defines what elements are included in the
system, what function each element has, and how each element
relates to one another. It is the big picture or overall structure of the
whole system—how everything works together. It follows that to
design a software system, a software architect has to take many
factors into consideration:
• the purpose of the system,
• the audience or users of the system,
• the qualities that are of most importance to users, and
• where the system will run.
Software architecture is important, particularly for large systems. If
there is a clear design of the overall system from the start, there is a
solid basis for developers to follow. Each developer will then know
what needs to be implemented and how things are related to meet
desired needs efficiently. This avoids conflicts, duplication, and ad
hoc unnecessary work.
Upon completion of this module, you will be able to:
(a) Explain the architecture overview and process.
(b) Describe Kruchten’s 4+1 View Model.
(c) Describe Component diagrams.
(d) Describe Package diagrams.
(e) Describe Deployment diagrams.
(f) Describe Activity diagrams.

Some advantages of software architecture include:
• higher productivity for the software team, as a well-defined
structure helps to coordinate work, implement individual
features, or guide discussions on potential issues
• improved evolution for the software, since design principles
are applied to make changes easier to accomplish or defects
easier to find
• enhanced quality in the software by carefully considering
the needs and perspectives of all the stakeholders
Software architecture helps make a system easier to maintain, reuse,
and adapt.
In order to develop software architecture, architects must take into
account the stakeholders of the system. Stakeholders are the
people who have an interest in the software at hand. They either use
the system or benefit from it in some way.

Software architecture typically have the following stakeholders.
Each will have their own perspective.
Stakeholder Description
Software
developers
Software architecture helps developers create
and evolve software by providing strong
direction and organization on what needs to be
done.
Project
Managers
Software architecture provides useful
information to project managers to help them
identify possible risks and to manage the
project successfully. Software architecture
helps project managers to understand task
dependencies and impacts of change and to
coordinate work assignments.
Clients Clients make important decisions about the
system, like its funding. Software architecture
establishes a basis for communication with
clients, so they understand what they are
paying for and that their needs are met.
End users Users may not care how the software is actually
designed, but they do care that it “works well”
for them.
One important way software architecture is presented is through
UML diagrams.
Kruchten’s 4+1 View Model
Multiple perspectives are necessary to capture the complete
behaviour and development of a software system. There are several
important considerations.
One consideration is the functionality of the software. Functionality
involves what a system does to satisfy the purpose the client desires.

Focusing on this functionality and the needed objects leads to a
perspective called the logical view.
Another consideration is how well the software executes, using
characteristics like the efficiency of the system or the interaction of
subprocesses. These characteristics affect the performance and
scalability of the system. Focusing on the processes implemented
by the objects in the logical view leads to a perspective called the
process view.
Software can also involve the development view. This perspective
focuses on implementation considerations such as the hierarchical
structure of the software. The programming languages of the
system will heavily influence this structure and therefore places
constraints upon development.
Another perspective of the software can be seen through the
physical view. The software will have physical components that
interact and need to be deployed. The interaction between these
different elements and their deployment will affect how the system
works.
All four views share the purpose or desired capabilities of the
software as defined by the client. Scenarios outline use cases or
tasks required by the end users, which provides context to help to
detail the four views.
Together, logical, process, development, and physical views, along
with scenarios form Philippe Kruchten’s 4+1 View Model. This
model is a way of understanding the key considerations or
important perspectives that need to be addressed in software
architecture.
Let us examine each of these views individually.
Logical View
The logical view, which focuses on the functional requirements of a
system, usually involves the objects of the system. From these
objects, a UML class diagram can be created to illustrate the logical
view.

A class diagram establishes the vocabulary of the problem and
resulting system. By defining all of the classes, their attributes, and
their behaviours it becomes easy to understand the key abstractions
and terminology. Class diagrams are also useful for specifying
database schemes. The class diagram makes it easier to see how
classes interact and how data should relate to each other in a
database.
Process View
The process view focuses on achieving non-functional
requirements. These are the requirements that specify the desired
qualities for the system, which include quality attributes such as
performance and availability. The process view also presents
processes that correspond to the objects in the logical view.
DID YOU KNOW?
Some of the most effective UML diagrams related to the logical
view of a system are the class diagram and the state diagram.
Both the class diagram and the state diagram focus on the
classes and objects of a system.
DID YOU KNOW?
Some of the most effective UML diagrams related to the process
view of a system are the activity diagram and the sequence
diagram. The activity diagram can illustrate the processes or
activities for a system. The sequence diagram shows how objects
interact with one another, which involves how methods are
executed and in what order.

Development View
The development view describes the hierarchical software structure.
It also considers elements such as programming language, libraries,
and toolsets. It is concerned with the details of software
development and what is involved to support that. This extends to
management details such as scheduling, budgets, and work
assignments. Essentially, the development view covers the
hierarchical software structure and project management.
Physical View
The physical view handles how elements in the logical, process, and
development views must be mapped to different nodes or
hardware for running the system.
Scenarios
Scenarios align with the use cases or user tasks of a system and
show how the four other views work together.
For each scenario, there is a script that describes the sequence of
interactions between objects and processes. This involves the key
objects defined in the logical view, the processes described in the
process view, the hierarchy identified in the development view, and
the different nodes specified in the physical view. Scenarios relate
these elements to provide a complete picture.
None of the views are fully independent of each other, with
elements of some views connected to others. The 4+1 view model
can be molded to fit many situations to understand the architecture
of a software system. Being able to see a complex problem in many
different perspectives helps make your software more versatile.
DID YOU KNOW?
One of the most effective UML diagrams related to the physical
view of a system is the deployment diagram. It can express how
the pieces of a system are deployed onto hardware or execution
environments.

Component Diagrams
UML component diagrams are concerned with the components of
a system. Components are the independent, encapsulated units
within a system. Each component provides an interface for other
components to interact with it. Component diagrams are used to
visualize how a system’s pieces interact and what relationships they
have among them.
Component diagrams are different from most other diagrams, as
they show high-level structure and not details like attributes and
methods. They are purely focused on components and their
interactions with each other.
Component diagrams are a static view of the software system, and
depict the system design at a specific point in its development and
evolution. The basis of component diagrams focuses on the
components and their relationships. Each component in a diagram
has a very specific relationship to the other components through the
interface it provides.
Component diagrams have ball
connectors, which represent a provided
interface. A provided interface shows
that a component offers an interface for
others to interact with it. The provided
interface means that client and
consumer components have a way of
communicating with that component.
Component diagrams also have socket
connectors that display a required interface.
The required interface is essential to the
component diagram, to show that a
component expects a certain interface. This
required interface is to be satisfied or
provided by some other component.

Finally, component diagrams can
illustrate an assembly relationship. An
assembly relationship occurs when one
component’s provided interface matches
another component’s required interface.
The provided interface is depicted by a
ball, and the required interface is
depicted by a socket.
To build a component diagram, first, you must identify the main
objects used in the system. Next, the relevant libraries for the
system need to be identified. Finally, the relationship between these
components would need to be identified.
When relevant libraries are identified, this extends to third-party
implementation dependencies, which should also be integrated
into the diagram where relevant. Below is an example of a
component diagram for a video game system.
Component diagrams are especially useful early in the design
process, because of its high-level emphasis. They can be drawn at
different levels and allows you to focus not only on systems but on
subsystems as well.

Package Diagrams
Sometimes, classes in object-oriented software are related in some
way.
A package groups together elements of software that are related.
Elements can be related based on data, classes, or user tasks. A
package can also define a “namespace” for elements it contains,
that is, a package is named and can organize the named elements
of software into a separate scope. An element can be uniquely
identified in the system by a fully “qualified name” that is based on
its own name and the name of the package that the element is in.
Package diagrams show packages and the dependencies between
them. These diagrams can organize a completed system into
packages of related packageable elements, which could include
data, classes, or even other packages. Package diagrams help
provide high-level groupings of a system so that it is easy to see
how a package contains related elements as well as how different
packages depend on each other.
Below is an example of a package diagram for a video game:

Packages are depicted by tabbed folders. If there are no elements
to show in the package, then the package name goes into the
centre of the folder. If details are needed, there are two ways this
can be expressed.
The elements can be nested within the folder, as below:
Alternately, this notation can be used:
Here, a player package contains the movement interface, but it is
drawn below the folder in the diagram. This relationship is a
composition, where the whole package has an element as a part.
Contained elements can also be listed in a package by their names.
These names can be partially qualified, but they should be unique
within the package.

Notice in the above example that plus (+) and minus (-) signs have
been used to indicate if elements are public or private. Public
elements can be accessed outside of the package by their fully
qualified names. Relationships are denoted through dotted-line
arrows.
A package can import an element from another package, as in the
example below where the Level package is importing Location from
the Player package. This brings the name Location into the Level
namespace, particularly for convenient use by elements in Level. As
the import tag is public, the name is visible for further import
through the Level package. If a tag was private, the name would
only be visible by elements within the Level namespace and not
further outside.
A package can import the entire contents of other packages, as in
the example below, where the User Interface package is doing a
public import of the Player package, as indicated by the import tag,
essentially doing public imports of the visible names of elements in
Player. The User Interface package is doing a private import of the
Buttons package, as indicated by the access tag, essentially doing
private imports of the visible names of elements in Buttons. These
names are not made further known outside the User Interface
namespace.

Packages can also be merged, as in the diagram below. Merging
typically occurs when two packages or concepts need to come
together into one. This is a use of generalization that allows different
definitions to be provided for the same concept.
The first diagram shown at the beginning of this section shows a
complete package diagram for a video game system. Packages
such as player, level, score, and analytics on player habits are
included. These packages mostly interact with each other, meaning
that they can publicly import each other. Note that the analytics
package, however, requires some communication with an online
server to store data. This package is available by private import.

The bonus relationship is also special because it has a uses
relationship with a normal level. The uses relationship implies that
the bonus level requires a normal level for its full implementation. If
there is no definition for level, there cannot be a definition for bonus
level.
Package diagrams can be created at any stage of development.
They can also adapt and change with the latest version of software
being worked on. Package diagrams are particularly useful for
technical designers because they allow them to see the
dependencies and relations between groups of related elements.
Deployment Diagram
UML deployment diagrams are used to visualize the deployment
details of a software system. The diagrams include more than just
code, but also separate libraries, an installer, configuration files, and
many other pieces. In order for software to be ready to run, it is
necessary to understand all the files and executables involved and
the environments where they reside.
The deployment environment, or deployment target, can be very
specific and involve particular hardware devices. It can also be very
general and involve supported operating systems. Details in a
deployment diagram change accordingly. For example, software
developed for Linux, MacOS or Windows may have differences from
one another.
Deployment diagrams deal with artifacts. Artifacts are a physical
result of the development process. Artifacts for a video game might
include things like an executable to run the game, an installer to
install the game, audio libraries for sound, and multimedia assets.
These are created as outcomes of producing the system and are the
final pieces to be put together.
There are two different types of deployment diagrams:
• specification-level diagrams
• instance level diagrams

A specification-level diagram provides an overview of artifacts and
deployment targets, without referencing specific details like
machine names. It focuses on a general overview of your
deployment rather than the specifics.
An instance-level diagram is a more specific approach that maps
specific artifacts to specific deployment targets. They can identify
specific machines and hardware devices. This approach is usually
used to highlight the differences in deployments among
development, staging, and release builds.
When creating deployment diagrams, it is important to use the
correct notation for the various elements. One of the most
important aspects of the deployment diagram is the node. Nodes
are deployment targets that contain artifacts available for execution.
In a deployment diagram, they look like 3D boxes.
Hardware devices are also displayed as 3D boxes, only they have a
“device” tag on them to differentiate them.
Relationships can be represented a few different ways in
deployment diagrams. A solid line between two nodes shows a
relationship between deployment targets. The line shows that the
two nodes have a communication path between them. This
relationship typically identifies a particular communication protocol.
If an artifact is drawn inside a node box, this shows that an artifact is
deployed to a node. This also means that the artifact cannot
function without this deployment target.

Manifestation is a relationship where an artifact is a physical
realization of a software component. It can be represented with a
“manifests” indicator, as below.
Here, the Player.class manifests the Player component, which is the
encapsulated unit that contains all of the functionality of a player.
Below is a specification-level deployment diagram for a simple
video game. This diagram provides a general and concise idea of
how to deploy the application without providing too much detail.

In a deployment diagram, there is a distinct hierarchy of
deployment targets. This hierarchy is very important. You should
start from the highest level of your deployment information, from
application name down to device and operating system. In this
example, the application, the device, the operating system, and the
execution environment are all listed. When you look at each of these
nodes, it is clear what environment is necessary to run the
application.
Note that only the most important artifacts are represented in this
diagram so that it is kept at a high level. Remember, deployment
diagrams provide a high-level overview of artifacts, libraries, main
components, machines, and devices that your application needs to
run.
Deployment diagrams help provide consistency and organization to
deployments, which helps avoid system failures. The diagrams also
help keep track of the files and executables needed to deploy and
run the software. The diagram can be on an instance level specific
to the deployment machines you are using, or it can be general for
a range of execution environments.

Activity Diagram
A UML activity diagram allows the representation of the control
flow from activity to another in a software system. It captures the
dynamic behaviour of the system and allows the mapping of
branching into alternative flows.
Activities are actions that further the flow of execution in a system.
They are actions that when completed cause another action to
execute. For example, an action can alter or create new objects.
These changes or actions can drive your application forward.
In order to create an activity diagram, you must:
1. Identify the activities.
2. Identify the respective conditions of the system’s activities.
Once activities and respective conditions have been determined,
the activity diagram can be created. There are a few major parts to
an activity diagram.
There are start and end notes that look like labelled circles. These
circles are where the diagram must begin. They show the starting
activity that initializes the control flow of the application. The end
node shows the final activity of the diagram. Intermediate activities
are shaped like an oval, and they describe all of the activities that
change the game state before the game ends.
Additionally, activity diagrams can include decision nodes. These
are diamonds that have an activity leading into it, and there is the
possibility of two alternative outcomes as the next activity. The
choice of outcome depends on how the condition on the decision
node evaluates. An example of a decision node can be seen below.

All essential activities must be included in a diagram as well as the
conditions.
Below is an example of an activity diagram that illustrates a video
game.

In this example, the activity diagram begins by playing the game.
The player then plays the level activity. When the level is complete,
the player encounters the first condition. If the score is high enough,
then the player may move on to the bonus level. If the player cannot
play the bonus level, then the game is ended, as denoted by a
circle.
Activity diagrams allow the mapping of concurrent activities that
happen in parallel. A fork in the flow moves into parallel flows, so
activities can happen at the same time. Parallel flows can join into a
single flow. For example, if a level ends in a video game, the
soundtrack flow will end at the same time as the play flow. To
denote this, activity diagrams can use a separate swimlane for each
flow.
Swimlanes can also divide activities up into different categories,
such as where it occurs or the user role involved. See the example
below for separating player activities and level activities.
In this diagram, there is a label for each swimlane. Arrows crossing
each lane show how different activities of a system interact across
categories.
Activity diagrams allow you to see what activities and conditions
should be included in a system. As well, from the diagrams you see
the order in which features are encountered while also allowing for
alternate flows to be taken into account for the system.

MODULE 2: ARCHITECTURAL STYLES
This module will explain important software architectures, their key
characteristics, and how they are used in practice. In particular, we
will examine language-based systems, repository-based systems,
layered systems, interpreter-based systems, dataflow systems,
implicit innovation systems, and feedback control systems.
Language-Based Systems
The programming paradigm of the language selected to implement
a system will affect the architectural style of that system. Each
programming paradigm has its own set of constructs, principles,
and design patterns, and their use shapes the system being created.
This section focuses on object-oriented architectural styles, which
result from object-oriented programming paradigms.
(a) Explain important software architectures, their key characteristics,
and how they are used in practice, which include:
• Language-based systems
• Repository-based systems
• Layered systems
• Interpreter-based systems
• Dataflow systems
• Implicit invocation systems
• Process control systems

In previous courses, the following object-oriented principles were
explored:
• Abstraction that simplifies a concept
• Encapsulation that bundles data and functions into a self-
contained object, so other objects can interact with it
through an exposed interface
• Decomposition that breaks a whole into parts
• Generalization that allows you to factor out conceptual
commonalities
Object-oriented design patterns were also explored. These can fall
into the following categories:
• Creational patterns that guide the creation of new objects
• Structural patterns that describe the relationships between
objects and the interactions between classes and subclasses
• Behavioural patterns that focus on how objects perform
work individually or as a group to accomplish something
These elements lead to an object-oriented architectural style for the
system.
Abstract Data Types and Object Oriented Design
Object-oriented design architectural styles are focused on the data.
When modelling a system in object-oriented design, begin by
looking at the different kinds of data handled by the system to see
how the system can be broken down into abstract data types. An
abstract data type can be represented as a class that you define to
organize data attributes in a meaningful way, so related attributes
are grouped together along with their associated methods.
Encapsulation restricts access to the data and dictates what can be
done with it.
Object-oriented refers to a system composed of objects where each
object is an instance of a class. In other words, the type of each
object is its class.

Objects may interact with each other through the use of their
methods. The object-oriented paradigm allows for inheritance
among abstract data types. This means that one abstract type can
be declared an extension of another type.
These classes themselves form a language-based architecture that
arises from basic object-oriented principles. The classes within the
system will determine the overall structure of the system. In other
words, the overall object-oriented architectural style of a system
directly follows from the fact that an object-oriented approach was
used in development.
Some problems are well suited to an object-oriented architectural
style. However, not all situations will have easily identifiable classes.
In some situations, another design choice may be better suited for
the problem at hand.
Instructor’s Note:
For a review on object-oriented thinking, see the Object-
Oriented Design course of Software Design and
Architecture specialization.
Main Program and Subroutine
Main program and subroutine architectural styles are focused on
functions. They develop from a procedural programming paradigm.
C is an example of a language that follows this paradigm.
Note that for this section, the terms routine, procedure, and function
are used interchangeably with the term subroutine.
In a main program and subroutine architectural style, a system is
modelled by breaking up the overall functionality of the system into
a main program and subroutines

The diagram above represents a main program and subroutine
architectural style. In the diagram, subroutines are connected by
procedure calls, and they may have nested calls. In nested calls,
subroutines may call other subroutines, which may call yet more
subroutines, and so on. This means that the structure of the
resulting code is not flat, but rather it is hierarchical, so it can be
modelled as a directed graph. The structure of the subroutines that
build up the system affect the structure of the system as a whole.
The subroutines declared in the code are structure as a big “call
tree.”
In this paradigm, data is stored as variables. Abstract data types are
supported in procedural programming; however, inheritance is not
explicitly supported. Under this paradigm, it is not easy to make one
abstract type an extension of another. The main focus of this
paradigm is therefore on the behaviour of functions and how data
moves through those functions. The main program and subroutine
architectural style is best suited for computation-focused systems.

Each subroutine may have its own local variables. A subroutine has
access to all data within its scope. To access data outside its scope,
data may be passed into the subroutines as parameters, by value or
by reference, and data may be passed out of a subroutine as a
return value.
A principle of this paradigm is “one entry, one exit per subroutine,”
which makes the control flow of the subroutine easier to follow.
This architectural style promotes modularity and function reuse,
which results in certain advantages. When functions are well defined
and do not produce side effects, they are considered like “black
boxes.” Given an input, they will always have the same output.
Further, library functions are easily integrated into programs.
There are also certain disadvantages from this style. Subroutines
may mutate data in unexpected ways. A subroutine may be affected
by data changes made by another subroutine during execution.
This issue is especially true for global data shared across
subroutines. Changes to data can be unpredictable and result in a
function receiving an input that it was not expecting, or even an
input that it is not capable of handling correctly, which can result in
run-time errors.
Procedural programming paradigms are well suited to systems
centred around computation, such as spending management
programs. Identifying object-oriented components of these kinds of
systems can be difficult and result in overly complex solutions.
However, for problems where abstract data types with modelling
make the solution easier, then object-oriented architectural styles
are more well suited.

Repository-Based Systems
Modern software systems rarely operate in isolated environments.
As a software developer, any software architecture you create needs
to be capable of sharing information between separate
components.
A number of issues need to be addressed to do this:
• The state of a component is only temporary while the
component is running. This means that any objects created
and any amount of data processing will be lost when the
component stops running.
• Components may not need all the data that is stored in the
files. There needs to be a way to communicate the state of
the data between multiple components.
Although it is possible to save information to keep in a file to
prevent the loss of data, files are not the best way to transfer data
between components.
To solve both of the issues, a data-centric software architecture
can be used. This architecture allows data to be stored and shared
between multiple components, thus increasing the maintainability,
reusability, and scalability of the system. This architecture can be
achieved by integrating a method of shared data storage, such as a
database, into the overall system design.
At the core of a data-centric architecture are two types of
components:
• Central data is the component used to store and serve data
across all components that connect to it.
• Data accessors are the components that connect to the
central data component. The data accessors make queries
and transactions against the information stored in the
database.

Below is a diagram of a data-centric software architecture.
Data accessors are separated from one another, and they
communicate only to the central data component. The central data
shares data by saving the desired information from the current state
of the system and serving data when requested.
In order to better understand this system, it is helpful to examine
each of these components.
Databases
Central data is often stored in a database. Databases ensure several
data qualities, which are important to data-centric architectures.
These are:
• Data integrity. A database ensures that data is accurate and
consistent over its lifespan.
• Data persistence. A database ensures that data will live on
after a process has been terminated—databases can save
data from any number of components.
One way that databases store information is by using tables.
Relational databases are a type of database that uses tables. Each
table represents an abstraction.

Relational databases use Structured Query Language (SQL) to
query or “ask” the database for information and to perform
transactions or “tell” the database to do something. These abilities
allow for information to be shared through a database between
data accessors.
The management and optimization of queries and transactions can
be automated by a database management system (DBMS), so
integrating a database into a system is simplified.
In data-centric architectural design, the central data is passive.
Central data is primarily concerned with storing and serving
information, not with heavy data processing or large amounts of
business logic.
Data accessors are any component that connects to a database,
which is characterized by its abilities to:
• Share a set of data, while being able to operate
independently.
• Communicate with the database through database queries
and transactions. Data accessors do not need to interact
directly with each other and therefore do not need to know
about each other.
• Query the database to obtain shared system information.
This is used to get data in order to perform computations.
• Save the new state of the system back into the database
using transactions. Data is stored back into the database
once the data accessor has finished its processing.
A data accessor contains all the business rules required to perform
its functions. This means that this software architecture enables you
to separate concerns into different, specialized data accessors. Also,
use of the data accessors can be controlled, so an end user only has
permission for the ones they need on a day-to-day basis.

Data-centric architecture presents many advantages over a basic
object-oriented system, because of the integration of a centralized
database. These advantages include:
• Increased support of data integrity, data backup, and data
restoration.
• Reduced overhead for data transfer between data accessor,
as data accessors do not need to be concerned with talking
to one another—the database communicates for them,
• A system that can be easily scaled up, as data accessors are
functionally independent, so additional features can be
added with having to worry about affecting others.
• Central data components usually “live” on a separate server
machine with sufficient disk storage dedicated to the
database, which allow for easier management of information.
Data-centric architecture also presents disadvantages. Integrating a
database can disadvantage a system in the following ways:
• The system is heavily reliant on the central data component,
so if it becomes unavailable or if the data corrupts, then the
entire system is affected. Safeguards such as data
redundancies that replicate data on separate hard disks can
be costly.
• Data accessors are dependent on what gets stored in the
database. New data accessors need to build around the
existing data schema. Anything that isn’t stored needs to be
computed, or if there is no matching column or table for a
specific data need, then the database cannot be used.
• It is difficult to change the existing data schema, particularly
if a large amount of data is already stored. Further, data
schema changes will affect data accessors.

In summary, data-centric software architecture is a commonly used
design by many companies. They allow large amounts of data to be
stored and managed in a central data repository, making a system
more stable, reusable, maintainable, and exhibit better
performance. They also separate the functionality of data accessors,
so it is easier to scale the system. Finally, they also facilitate data
sharing between data accessors through database queries and
transactions.
As always, the decision to use a data-centric architecture relies on
the context of the problem at hand.
Layered Systems
Many organizations are designed by hierarchy. Consider a school,
for example, which could be visualized as below:
In this diagram, there are three groups, which are the three layers of
the organization. People in the adjacent layers interact. So, students
and the principal only interact with one layer—the teachers, while
teachers interact with two adjacent layers—students and the
principal. People in each layer can also interact with each other. So,
teachers can also interact with other teachers.
In software, a layer is a collection of components that work together
towards a common purpose. The key characteristic of a layered
architecture is that the components in a layer only interact with
components in their own layer or adjacent layers. When an upper
layer interacts with the layer directly below it, it cannot see any of
the deeper layers.

Usually, the inner or bottom layer provides services or abstractions
for the layer outside or above it. The interfaces provided by the
components of a layer should be well defined and driven by the
needs of the system.
Layering allows for applying “separation of concerns” into each of
the layers. Many layered systems are split into “presentation,”
“logic,” and “data” layers.
An operating system for a computer is a common example of a
layered system.
The kernel is the core of the operating system. One of its main
responsibilities is to interface with the hardware and allocating
resources. Above the kernel layer is the system and application
libraries layer. This layer provides the higher-level functions for
saving files or drawing graphics. Above this layer is the utilities and
applications layer, which to most users is the operating system.
Utilities are tools included with the operating system, such as
command-line programs. Applications are additional programs,
often installed by the user for specific needs.

Advantages of layered systems include the fact that:
• Users can perform complex tasks without needing to
understand the layers below.
• Different layers can be run at different levels of authorization
or privilege. Typically, the top layer is considered the “user
space” that does not have authority to allocate system
resources or communicate with hardware directly. This
“sandboxing” provides security and reliability to the kernel.
• Designs will be more loosely coupled, as layered
architecture follows the principle of least knowledge.
Lower layers typically provide services to the layer above. If a layer is
replaced, the provided interface for the layer above must be
consistent with the previous implementation.
Not every software problem can easily be mapped to a layered
architecture. It may not always be obvious how a solution can be
made into layers and where the division for software will be for a
collection of object-oriented classes.
There are certain trade-offs to be aware of in layered systems. Most
notably, enforcing layers has an efficiency trade-off. If only adjacent
layers can communicate, then the system will likely have some
interactions that pass through one layer to the next, and sometimes
information must be shared between layers that are not adjacent.
This extra communication adds complexity and uses up processing
resources.
Overhead must be balanced against the separation of concerns
gained from enforcing layers. If most communication is passing
directly through a layer, then the design may need to be relaxed, or
a different architecture may be better suited to the problem. If only
a small amount of communication is passed through, then a layered
system may be a good choice.
Layered architecture can be relaxed by allowing for passthrough. In
a diagram, this is shown as a notch in the layer. Let us return to the
example at the beginning of this section—a school organization.
Suppose that the principal would like to speak to the students.

The diagram would be reworked as below:
This is another trade-off, however. Although communication has
opened up, the organization is now more tightly coupled, and more
complex.
Some layered systems may have more than one grouping in a layer.

In summary, layered architecture:
• It can be intuitive and powerful. Many organizations and
solutions have layered structures, so it is easy to apply
layered architecture where appropriate.
• It supports the separation of concerns, because each layer is
a set of components with similar responsibility or purpose.
• It is modular and loosely coupled, as each layer only
communications with one or two other layers, allowing for
different implementations to be easily swapped.
• It can be adapted so that layering is not always strict, which
helps manage design complexity or provide a starting point
for structuring the system.
Client Server n-Tier
n-Tier or multitier architectures are related to layered architectures.
Tiers refer to components that are typically on different physical
machines. Although the terms “tier” and “layer” are used
interchangeably, they are not the same thing. n-Tier or multitier
architectures are layered architectures based on tiers.
The number of tiers in an n-Tier architecture can vary, although
three-tier and four-tier architectures are common. The relationship
between two adjacent tiers in an n-Tier architecture is often a
client/server relationship. In a client/server relationship, one
program (the server) provides services such as storing information
in a database or performing computation tasks upon request from
another program (the client). This communication is known as the
request-response.
A tier can act as both a server and a client, fulfilling the requests of
its clients and making requests of its servers at the same time.

Consider the deployment diagram below, illustrating the setup of a
simple media server.
One computer hosts the media server process, where movies and
TV shows are stored and can be streamed on demand. When a
client requests a movie or TV show, the media server streams it to
the client. The server process does not know the number of clients it
may have. More clients can be added as necessary.
In this example, server and client processes are on different
computers; however, they can run on the same machine. The media
server can be hosted on computer and use either a media player on
the same desktop or a mobile device such as a tablet to watch the
media.
Client-host and server-host refer to machines that host client
software and server software respectively. In this diagram, the top
row illustrates client-hosts, and the bottom row illustrates server-
hosts. This diagram is therefore a two-tier architecture. The client-
hosts make up the client tier, and the media server is the data tier.
A server is usually specified by the type of service it
provides. For example, these servers might be called web
servers, application servers, file servers, and print servers.
The request-response relationship from this example can also be
illustrated in a sequence diagram.

The client, in this case, a media player, requests to see what the
media server has in its library. The server sends that information
back. After the user makes a selection, the client requests the
chosen video from the media server, which starts streaming it.
Request-response relationships can be synchronous or
asynchronous.
• A synchronous request-response relationship is one where
the client sends a request, then awaits the server’s response
before continuing execution. In UML, synchronous messages
are depicted with a closed arrowhead. Servers do not always
respond quickly, so a synchronous message may cause the
client user interface to freeze while waiting for a response.
• An asynchronous request-response relationship is one
where the client sends a request, but control returns right
away, so it can continue its processing on another need.
None of this processing can depend on the response from
the server. Once the server has completed the request, a
message is sent to the client, which will have a handler to
process the response. In UML, an asynchronous message is
depicted with a line-arrowhead.

Client requests are often designed to run asynchronously, because
it is often preferable to have clients be responsive for other tasks
even if they are waiting for information from the server.
Limiting client/server relationships to request/response messaging
patterns allows for systems that can be scaled more easily by adding
clients. Clients and servers are extensible, because as long as a
server receives a message it knows, it will respond. The source of
the message is not important to the server, so many clients can be
added as needed.
n-Tier architecture can take on different numbers of tiers, based on
separation of concerns. Additional tiers can be added to the system
for specific purposes. The example above, of a simple client-server
relationship with one server and one set of clients, is known as a
two-tier architecture.
However, you could also build a three-tier architecture, where a tier
is inserted between the database and end users. This tier is often
referred to as a middle layer, a business layer, or an application
layer—its name depends on its main responsibility. The middle layer
will be a client of the database, and a server for the client
application software on the end users’ devices. Middle tiers help
determine how or when data can be changed, and in what ways.
Having a middle layer allows client application software to be
thinner, as it can be focused on presentation only. This makes it
easier to maintain.
You could add even more tiers to your architecture if you want, as
long as the tiers serve a specific relationship.

Two potential drawbacks of n-Tier architecture include:
• It requires extra resources to manage the client/server
relationships. If more tiers are added, there are more
machines or processes to manage, with different
communication protocols. This makes a system more
complex, and therefore more difficult to change and
maintain.
• A server acts as a central point of failure. Systems may have
backups or mirrors, but it can take time to switch to these
backups and recover the server. Redundant servers are
possible but add complexity.
The advantages of n-Tier architecture are notable, however, and
contribute to its common use in the software industry.
• n-Tier architecture is very scalable. Clients can continue to be
added as long as a server can handle all the requests it
receives in a reasonable time.
• Centralization of functionality allow for data to reside on one
machine but to be accessible by any machine on the same
network.
• Centralization of computing power allows client machines to
require less processing power. Companies can thus offer
processing power as a service, which is more practical and
cost effective.
• n-Tier architecture supports separation of concerns. Middle
layers can take the role of managing application logic and
accessing the database directly. Adding tiers can further
allow for loose coupling and levels of abstraction, making a
system that is easier to change and extend.
When a system can be split into service roles and request roles, a
client/server architecture should be considered.

Interpreter-Based Systems
In some software systems, users can write scripts, macros, or rules
that access and compose the basic features of the system in new
ways.
For example, users can write formulas in Microsoft Excel. A
component known as an interpreter is built into the system that
allows Excel to carry out the calculation in the formula “at run time”.
The interpreter “interprets” the formula into an intermediate
representation and “executes” the representation. The end user
does not need to know about the underlying implementation of the
function. Systems use interpreters to drive programmable actions
specified by the user.
Interpreter-based systems allow end users to write scripts, macros,
or rules that access, compose, and run the basic features of those
systems in new and dynamic ways. It provides users with flexible
and portable functionality that can be applicable in a variety of
commercial systems.
Interpreters can run scripts and macros. Scripts are often used for
automating common tasks, such as scheduling tasks, performing
repetitive actions, and composing complex tasks that invoke other
commands. Macros are an evolution of scripts that became popular
with the introduction of graphical user interfaces. A macro records
keyboard and mouse inputs, so they may be executed later.
With an interpreter, end users can add functionality to a system and
extend existing functionality of a system, by composing pre-existing
functions together in a specific sequence, in order to create
something new. These pre-existing functions are defined by the
system architecture and offered to the user. The developer does not
need to implement all possible combinations of functionality.
Interpreter-based systems encourage end users to implement their
own customizations. This is further supported by systems that offer
an easier-to-use language with domain-specific abstractions suited
to the needs and thinking of the end users. End users do not have to
know the programming languages used in developing the software
in order to customize the system for their needs.

Interpreter-based systems offer certain advantages. In particular,
interpreters make systems more portable, so they can work on
platforms that the interpreter supports. This is an important feature
with the growth of virtual machines and virtual environments—more
and more services are being hosted in the cloud.
However, interpreter-based systems also offer certain
disadvantages. Interpreters can be slow. Basic implementations
spend little time analyzing the source code and use a line-by-line
translate and execute strategy. This is a trade-off: it may be faster
and more flexible for developers and end users to use an
interpreted language, but slower for computers to execute
interpreted code.
DID YOU KNOW?
Java also uses interpreters!
In Java, programs are first translated into an intermediate language
that is loaded into a Java Virtual Machine (JVM), which executes the
intermediate language. The JVM will attempt to optimize the
intermediate instructions by monitoring the frequency at which
instructions are executed. The instructions that are executed frequently
get translated into machine code and executed immediately.
On the next execution of the same intermediate instructions, the JVM
uses lazy linking to point the program to the previous machine code
translation. Instructions that are not used frequently are left for the
interpreter of the JVM to execute.
This decreases execution time since frequently used instructions do not
need to be constantly translated and the entire program does not need
to be translated all at once. The JVM also provides portability to Java
programs, allowing them to run on many operating environments.

Dataflow Systems
Data flow systems involve a series of transformation on sets of
data. Data is transformed from one form into another using different
types of operations.
This section will explore designing a data flow system through the
pipe and filter architectural style, a specific type of data flow
architecture.
Pipes and Filters
A pipe and filter architectural style has independent entities called
filters, which perform transformations on data input they receive,
and pipes, which serve as connectors for the stream of data being
transformed.
Below is a simple diagram outlining how a pipe and filter
architecture works.
Data flows in one direction, beginning at the data source and
moving through a series of pipes and filters, to end at the data
target. Changes to data are done sequentially from filter to filter.
Each pipe serves as a connector for the stream of data. It transmits
the output of one filter as input for the next filter. Each filter
performs its local transformation on the data input it receives from
the previous pipe and then sends the output on to the next pipe.
The order in which the filters transform data may change the end
result. This is similar to mathematics—the order in which addition or
multiplication occurs in an expression can change the end result.

UML component diagrams are a good option for showing the
details of pipe and filter architecture.
The data source can be connected to a number of different filters.
Each filter has a required interface to receive input from data
sources and other filters. They also have a provided interface that
will output transformed data. The data target is connected to
receive the fully transformed result.
Advantages to using the pipe and filter architecture include:
• Loose and flexible coupling of components. In particular, this
applies to filters. Each filter runs independently of other
filters, which allows the easy addition of new filters, of
moving filters around, or of changing individual filters in a
system. ‘
• Filters can be treated as black boxes. Users of the system do
not need to know the inner workings of each filter.
• Reusability. Each filter can be called and used over and over
again with different inputs. Filters can be repeatedly used in
many different applications for different purposes.
Disadvantages to using the pipe and filter architecture include:
• Reduced performance due to excessive overheads in filters.
Each filter must receive input, parse that input into some
data structure, perform transformations, and send data out. If
every filter must do this, it is a lot of overhead, and the
performance of the system will be affected as more and
more filters are added.
• Filters may get overloaded with massive amounts of data to
process. If one filter is working to process large amounts of
data, other filters may be waiting for inputs.

• This architecture cannot be used for interactive applications.
Applications that require rapid responses will not find this
architecture suitable, as data transfers and transformations
take time depending on the amount of data being
transmitted.
Pipe and filter architectures are popular in use, such as in the UNIX
operating system for text-based utilities. If different data sets need
to be manipulated in a system, the pipe and filter architecture is a
good choice.
Implicit Invocation Systems
In implicit invocation systems, functions are not in direct
communication with each other. The event-based architectural
style is one design for such a system. This style derives from the
event-driven programming paradigm.
Event Based
In event-based architectural styles, the fundamental elements in the
system are events. Events are both indicators of change in the
system and triggers to functions. Events can be signals, user inputs,
messages, or data from other functions or programs.
Under the event-driven programming paradigm, functions take the
form of event generators and even consumers. Event generators
send events, and event consumers receive and process these
events. A function can be both an event generator and an event
consumer.
In the event-based architectural style, functions are not explicitly
called. Instead, event consumers are based on events sent from
event generators. Event-based functions experience implicit
invocation, where communication between functions is mediated
by an event bus. An event bus is the connector between all event
generators and consumers in the system.
To achieve this structure, an event and an event consumer must be
bound via an event bus. This means that each event consumer
registers with the event bus to be notified of certain events.

When the event bus detects an event, it distributes the event to all
appropriate event consumers. The event bus is a critical part of the
system. The observer design pattern actually manifests the event-
based architectural style.
One means of implementing the event bus is to structure the system
to have a main loop that continually listens for events. When an
event is detected, the loop calls all the functions bound to that
event. This function must likely also be an event generator, because
often an event consumer will have to notify other functions when it
has completed its task or to send a state change. This function will
implicitly invoke other functions to run after it has completed. It
does this by sending out an event to the event bus. When an event
reaches the event bus, corresponding event consumers will be
triggered and the next computation can take place.
When designing your system, it is important to take into account
that indirect communication between functions may not be as
efficient as explicit function invocation. In event- based architectural
styles, event generators do not necessarily know which functions will
be consuming their events, and likewise, event consumers do not
necessarily know who is generating the events they handle. This is
called loose coupling, which makes it easier for systems to evolve
and scale up—adding a new function requires only registering a new
event-function pair to the event bus and adding a new event
consumer.
Updating objects and data in an event-based system must also be
considered.
Event consumers can change shared states as they process an event
if objects in the system are globally accessible. However, this allows
event consumers can be called asynchronously, that is, an event
consumer does not need to wait for other event consumers to finish
running before running itself. Two event consumers could be
running at the same time on shared data.
This asynchronous design may increase efficiency of the system, but
it can also result in race conditions. Race conditions mean that the
correct behaviours of functions is sensitive to timing or order, so
shared data may not be updated correctly.

Functions can be coordinated for access to shared data through a
semaphore. A semaphore is a variable or abstract data type that
toggles between two values, available and unavailable, and that is
used to control access to shared data. Available indicates that the
shared data is not in use, and unavailable indicates that the shared
data is in use by a function.
In general, a semaphore has a special operation to check and
toggle its value in one step. Any function that would like to access
the shared data must first check the value of the semaphore. If the
value is set to unavailable, then the function must wait. If the value is
set to available, the function can access the shared data, and the
value is set to unavailable. When the function no longer needs
access, it will toggle the semaphore value to available.
If an object is not globally accessible and its state is changed by an
event consumer, then some indication must be sent back to the
event bus before the function ends. Event consumers can thus be
aware of the state change to that event.
In event-based architectural styles, events and functions do not
behave in a predictable way. There is no guarantee of exactly when
an event will be handled, how long it will take to handle, or when an
event generator will emit an event. Control flow is thus based on
which events occur during its execution and in what order.
Event-based architecture styles are well suited to interactive
applications. Graphical user interface applications that rely on user
input and distributed systems that interact with other programs are
prime candidates for this style.
Cookie Clicker Example
Let us examine event-based architecture through an interactive
game, “Cookie Clicker.” The goal of this game is to collect as many
points as possible by clicking on an image of a cookie with your
cursor. This can be done manually or by clicking on a cursor icon.

This icon has a functionality—it will purchase a special automatic
clicker that automatically clicks the cookie at regular time intervals,
thus reducing work on the user’s part to collect points. Users can
“purchase” multiple cursors at the cost of some points.
Examine the component diagram below to see how this can be
designed with an event-based architecture.
A main event loop listens for event. Depending on where users click
within the window, a different event consumer will be invoked.
When the cookie is clicked, the function registered to handle this
click event is called and total points increases by one. When the
“buy clicker” icon is clicked, then the function registered to handle
this click event is called. This first reduces the score, as a purchase
was made. Then, an automatic clicker function is added to the
system, along with a timer function.
The timer function is an event generator, registered to the event
bus, that sends a “timer event” every five seconds. The event bus
detects when such an event is emitted and triggers every automatic
clicker function to consume this event. Each automatic clicker
function is responsible for making one click to the cookie. Every
time a timer event is sent to the event bus and received by the
automatic clicker functions, the total points increase by the number
of automatic clicker functions. These event consumers, event
generators, and event bus form the system of the game.

Process Control Systems
Process control is important in many processes to ensure efficient
and safe operations.
Feedback Loops
A feedback loop is one of the most basic forms of process control.
It has four basic components: a sensor, a controller, an actuator, and
the process itself. The sensor monitors some kind of important
information. The controller is the logic of the system. The actuator is
the physical method of manipulating the process. The process is
what is trying to be controlled.
For example, consider controlling room temperature. This is a
feedback loop. A thermostat acts as a sensor for the room’s
temperature. The software is the controller. The heating vent in the
room is the physical means to manipulating the process. The room
temperature is the process.
In a feedback look, the controller logic must run continuously. This
is because the process is always changing.
The frequency, or how often the loop runs, depends on the desired
level of control and the sensitivity of the system. For example, room
temperature changes slower compared to other processes, so it is
not necessary or useful to have a high frequency, such as every
microsecond.
Feedback loops can often be improved. Consider the example
above of room temperature. Let us improve that loop by
introducing a proportional controller. A proportional controller
calculates an error value as the difference between the desired
setpoint and a measured process variable and then applies a
correction based on proportion. In this case, a heating vent whose
position can be controlled could read an error, and open the vent
according to how big the error signal is. As the temperature comes
closer to the setpoint, the heating will slow down so that the
setpoint can be approached gradually. (Proportional controllers can
also have undesirable effects, but this runs outside the scope of this
course.)

Other Variations
Other variations exist on process control.
A feedback loop is an example of a closed loop, as information from
the process is used to control the same process. An open loop is
one where the process is controlled without monitoring the process.
Open loops cannot deal with changes in the process or check itself
to see if it is succeeding.
Feedforward control occurs in systems with processes in series.
Information from an upstream process can be used to control a
downstream process. Feedforward control requires a good model
of how process response and deals with unknown events. These
systems are often used in conjunction with feedback loops.
For example, a flood protection system could use feedforward
control. If an upstream monitoring station detects large flow rates,
then the controller for the flood protection system could be
signalled. The controller asks the actuator, a gate, to open and
divert water into a reservoir to keep water levels manageable.
More complex problems will have many sensors and ways to control
the process. Process control architecture for complex problems will
therefore be more complex. Below is a diagram for a complex
example of a self-driving car.

This architecture is based on a MAPE-K structure, which has four
major steps: monitor, analyze, plan, and execute. All four of these
steps must have knowledge of the process.
Step Description
Monitor The component or collection of components that
interfaces with sensors, turning raw signals into
meaningful information for the software.
Analyze The software must next analyze this data. Data could be
integrated from several sensors, so analysis might be
quite sophisticated. More sophistication generally
requires a better model of the process.
Plan Next, planning needs to be done. More advanced
planning needs will require more knowledge. This
knowledge is in the form of a model of the system. Such
models should be designed in close conjunction with
engineers specializing in the process.
Execute The final step is to execute. Analysis and planning
should make execution relatively simple. They will
simply turn desired actuator positions into signals to
communicate to those actuators.

Knowledge is very important in MAPE-K systems, particularly
complex ones. Currently, technology companies are investing
heavily in such systems, due to the advent of machine learning and
big data. This investment goes towards real-world testing, machine
learning, and data modelling in order to operate these systems.
Effective software architecture is key. Even though it is handled by
specialized professionals, it is important to be aware of process
control. Software is becoming a bigger part of controlling physical
systems.

MODULE 3: ARCHITECTURE IN PRACTICE
This module will focus on how software architecture is used in
practice. This includes exploring how software architectures are
analyzed, evaluated, and how they interact with software
development. This module also explores what makes an
architecture good or bad for a given problem, how architecture is
evaluated for quality, how to identify improvements in software
architecture, how to use architecture to plan development and
structure a development team, and how to identify opportunities for
reuse of architecture to reduce costs and increase development
efficiency.
Quality Attributes
Software architecture aims to combine software design patterns and
principles in order to define the software’s elements, their
properties, and how the elements interact with each other. This
suggests that these elements alone can determine the capabilities
and the quality of software architecture. However, as design
patterns address only a specific problem and not all business needs,
this is not the case. Modern systems need to be able to focus on a
wide range of problems, not just technical issues.
(a) Analyse, evaluate, and describe how software
architectures interact with software development.
(b) Identify what makes an architecture good or bad for a
given problem.
(c) Evaluate software architecture quality, and identify
improvements.
(d) Understand how architecture can be used to plan
development and structure a development team.
(e) Identify opportunities for reuse to reduce costs and
increase development efficiency.

Individual design patterns and principles are concerned primarily
with functional software issues, while software architecture
considers functional and non-functional requirements. Software
architecture must address the big picture. Architecture sets
guidelines for design patterns and principles in order to ensure
conceptual integrity and consistency throughout.
For example, design patterns covered in the previous course do not
touch a number of software qualities, such as testability, usability,
and availability. Software architecture addresses these qualities by
carefully structuring and coordinating design patterns in order to
construct a unified system. Software architecture is qualified by how
well the “entire” combination of software elements work together to
provide functionality and handle potential system issues as well as
how well the design addresses user experience and ease of
development.
Software architecture does not have inherent qualities that make it
“good architecture” or “bad architecture.” Instead, software
architecture is designed to address a set of requirements.
Requirements are used to address a problem or need. Further,
different architectures are meant to be used in different
environments and contexts. This means that an architecture isn’t
bad in some cases, but rather just be incorrect for the context it is
being used in.
Most systems use a combination of architectural designs. The
requirements for modern systems are complex, and there may not
be a clear-cut software architecture that can address all of them.
This is particularly true for non-functional requirements, which are
not always clear or explicitly presented by clients and stakeholders.
Non-functional requirements can also vary between different
stakeholder groups. For example, an end user may not care about
testability, while a development group will.
Software architecture should address the concerns of all
stakeholders, however. Prioritize system requirements and don’t be
afraid to make compromises in order to find a good balance
between system qualities.

If system architectures are not measured as inherently good or bad,
there must be a way to determine if the architectural design is
capable of meeting system requirements. This quality of a system is
determined by quality attributes. Quality attributes are measurable
properties of a system used to gauge a system’s design, run-time
performance, and usability.
For an attribute to be measurable, there needs to be an objective
means of quantifying it. For example, system availability can be
measured by the system’s uptime in some unit of time.
This section presents industry consensus for quality attributes.
However, these can be added or removed from your evaluation of a
system, based on specific needs. As always, context is important in
evaluating what is best for your system architecture.
This specialization has emphasized maintainability, reusability,
and flexibility when designing and implementing systems. These
attributes will be covered, as well as other important attributes from
a developer perspective, as outlined in the table on the following
page.

Quality
Attribute
Measurement of
Maintainability The ease at which your system is capable of
undergoing changes.
Systems will undergo changes throughout its life
cycle, so a system should be able to accommodate
these changes with ease.
Reusability The extent in which functions or parts of your system
can be used in another.
Reusability helps reduce the cost of re-implementing
something that has already been done.
Flexibility How well a system can adapt to requirements
change.
A highly flexible system can adapt to future
requirements changes in a timely and cost-efficient
manner.
Modifiability The ability of a system to cope with changes,
incorporate new, or remove existing functionality.
This is the most expensive design quality to achieve,
so cost of implementing changes must be balanced
with this attribute.
Testability How easy it is to demonstrate errors through
executable tests.
Systems should be tested as tests can be done
quickly, easily, and do not require a user interface.
This will help identify faults so that they might be
fixed before system release.
Conceptual
Integrity
The consistency across the entire system, such as
following naming and coding conventions.

In addition to qualities to account for from a developer’s
perspective, it is also necessary to take into consideration qualities
from a user’s perspective. The table below covers some of these
quality attributes and what they are measuring.
Quality
Attribute
Measurement of
Availability The amount of time the system is operational over a
set period of time.
A system’s availability is measured by uptime, so it
can be determined how well the system recovers from
issues such as system errors, high loads, or updates. A
system should be able to prevent these issues from
causing downtime.
Interoperability The ability of your system to understand
communications and share data with external
systems.
This includes the system’s ability to understand
interfaces and use them to exchange information
under specific conditions with those external systems.
This includes communication protocols, data formats,
and with whom the system can exchange information.
Most modern systems are interoperable and do not
exist in isolation.
Security The system’s ability to protect sensitive data from
unauthorized and unauthenticated use.
This attribute is important as most systems generally
contain sensitive information. This information should
be protected from those not authorized to see it, but
readily available to those who have authorization.
Closely associated with this is data integrity. A system
should be able to control who can see the data versus
who can change it.

Performance The system’s throughput and latency in response to
user commands and events.
Performance determines how well your system is able
to respond to a user command or system event.
Throughput is the amount of output produced over a
period of time. Latency measures the time it takes to
produce an output after receiving an input. This
attribute has a major influence on architecture due to
its importance to users.
It is better to have a lower price per unit of
performance. With the advancements in technology,
this ratio has been steadily decreasing.
Usability The ease at which the system’s functions can be
learned and used by the end users.
Usability determines how well your system is able to
address the requirements of end users. In order to
have high usability, your system needs to be easy and
intuitive to learn, minimize user errors, provide
feedback to the user to indicate that the system has
registered their actions, and make it easy for the user
to complete their task.
As software is not inherently good or bad, it is very important to
create or choose an appropriate architectural design for your
systems. The system architecture will ultimately determine how
functionality is implemented, how subsystems communicate with
each other, and how end users will interact with the system.
Software architecture is also important, as it makes maintaining,
supporting, and updating the system throughout its life cycle much
easier. If the qualities of a system are poor, then it will be difficult to
support changing needs.

A high-quality system does not need to be “complex.” An overly
complex system makes it difficult and time consuming to produce. It
is good practice to try to minimize complexity in the design. If a
simpler architecture can satisfy all the system requirements and
achieve a high-quality design while needing less time and less
money, it makes sense to go with that.
It is important to have detailed and up-to-date documentation of a
system. In fact, this is part of a high-quality system. This helps record
and share an architectural vision. The documentation keeps a
design cohesive, should the architect or any of the developers leave
the team.
Creating a high-quality system architecture should be methodical
and use a set of rules or guidelines for the design process. Although
every company will have some specific design rules or system
structure rules, some should be common sense.
Design process guidelines include:
• Recognizing the important of quality attributes and
prioritizing them for each system being designed. Quality
attributes should be kept up to date throughout the life cycle
of the system. Trade-offs and details of how a system will
meet qualities should be noted.
• Involving a technical lead in the design process. Although
architectural design can be applied to many different
technologies, involving a technical lead will help identify any
implementations that may pose a challenge, which may need
to be re-considered in the design.
• Taking a design approach from the perspective of different
groups of stakeholders.

Structural rules that ensure there is conceptual integrity when
implementing the system include:
• Having well-designed subsystems that are assigned
responsibilities based on design principles. Remember that
subsystems should use design principles such as separation
of concerns.
• Having consistent implementation of functions across the
entire system. If the same task needs to be achieved by two
or more subsystems, then consider how implementation can
be reused.
• Having a set of rules on how resources are used. Resources
may include memory, bandwidth, or threads that the system
has access to.
In summary, when selecting the appropriate architecture for the
problem at hand, it is important to involve all groups of
stakeholders in the design of the system, to adopt good
documentation practices, and to set rules for design and
implementation.
A well-designed system considers quality attributes from a
developer’s perspective, which includes maintainability, reusability,
flexibility, modifiability, testability, and conceptual integrity; the
system should also consider attributes from a user’s perspective,
which includes availability, interoperability, security, performance,
and usability.
Analyzing and Evaluating an Architecture
In addition to designing a system, it is important to know how to
evaluate the design to determine if it addresses the concerns, or the
requirements, of all stakeholders. Analyzing and evaluating software
architecture can be difficult due to the abstract nature of software.
It is important to methodically analyze and evaluate a system’s
behaviours, quality attributes, and various characteristics. This
section will look at how individual qualities are specified for
evaluation, and how to analyze the architecture as a whole.

In order to measure quality attributes, quality attribute scenarios
can be used to determine if a system is able to meet requirements
set for the attribute. There are two kinds of scenarios:
• A general scenario, which is used to characterize any system
• A concrete scenario, which is used to characterize a specific
system
Both general and concrete scenarios have:
• A stimulus source, which is anything that creates a stimulus. It
can be internal or external to the system.
• A stimulus is a condition that will cause the system to
respond. As the source of the condition can originate
internally or externally, types of conditions will need to be
differentiated.
• An environment is the mode of the system when it is
receiving a stimulus. This is an important aspect, particularly
if the system involves distributed computing, or if it can exist
in operational modes besides just running and stopped, like
recovering from a failure.
• An artifact is the part of the system that is affected by the
stimulus. In large-scale systems, a stimulus should not
directly affect the entire system.
• A response is how the artifact will behave as a result of
receiving a stimulus. This response could include handling
an error, recovering from a failure, updating system logs,
dispatching security alerts, or changing the current
environments.
• A response measure is a metric used to quantify the
response so that the quality attribute can be measured. The
metric should be quantitative and objective, such as
probability of failure, response time, repair time, and
average system load.

Scenarios are built to identify situations that impact the quality
attributes of a system. In the context of analyzing and evaluating
architecture, you should focus on situations that are outside of the
normal execution path. This means that scenarios involving
incorrect input, heavy system loads, or potential security breaches
should be prioritized highly.
It is a system’s “inability” to handle unexpected failures that stops it
from achieving a specific quality attribute.
Availability Example
Let us consider scenarios through an example. Imagine you are
addressing the availability of a system. In addition to focusing on
when a system is online and behaving normally, you have to
consider situations where the system becomes unavailable.
In a general scenario, high-level events are summarized. As a single
scenario may involve many component values, it often summarized
in a table.

Below is an example of a general availability scenario.
Scenario
Component
Scenario Component Value
Stimulus Source - End user
- Internal subsystems
- External subsystems
Stimulus - Incorrect user input
- Internal exceptions
- Unrecognized system request
- High request volume
- Heavy system load
Environment - Normal operating environment
- Starting up
- Shutting down
- Heavy load operating environment
- Recovering from error
- Processing request
Artifact - System servers
- System process
Response - Log exceptions
- Notify user
- Send error response to external system
- Redistribute data processing
- Redistribute system requests
- Notify user and external systems that the
system is shutting down/starting up
Response
Measure
- Time to restart (shutdown and startup
sequence)
- Time to undergo recovery
- Time to complete requests when the
request volume is high
- Time to complete a process under heavy
system load
- Time to become available after
encountering an incorrect input or request
Concrete scenarios are more focused. They can help you test an
architecture with a specific stimulus under specific system
environments and measure how well the system can respond.

Let us expand our example on availability. Availability of a web
server can be hindered in its ability to process requests when at
resource limits or under heavy load. A server’s availability should be
measured under different conditions.
Consider the specific concrete scenario below, where a customer
needs to wait for the system to finish processing previous orders for
concert tickets before they are able to send their purchase request.
The system is unavailable for customers if the system can’t accept
their request for tickets, which has a negative effect on the
availability quality attribute of the system.
Architecture Trade-off Analysis Method
There are a number of methods that can be used to implement
scenarios in analysis and evaluation of the entire architecture. This
section will focus on an evaluation technique developed by the
Software Engineering Institution at Carnegie Mellon University
known as Architecture Trade-off Analysis Method (ATAM).
With ATAM, evaluators do not need to be familiar with the
architecture or the problem space. A system can be evaluated by
outsiders.

There are three different groups of participants in ATAM.
• The first is the evaluation team. This has three different
subgroups: designers, peers, and outsiders. Evaluation
teams must be unbiased.
o Designers are the subgroup involved with
architectural design. They naturally follow
iterative, hypothesis-driven methods when
designing, including analyzing systems
requirements, creating a design to address those
requirements, and reviewing the design.
o Peers are the subgroup composed of those who
are part of the project, but they are not involved in
the design process. Their point-of-views helps
round out design decision.
o Outsiders are the subgroup composed of those
external to the project or organization.
Involvement of outsiders helps eliminate bias
towards the project in evaluation. Outsiders
should have experience and expertise in
analyzing architecture.
• Project decision makers are a participant group of
project representatives with the authority to make project
decisions. This could include project managers, clients,
product owners, software architects, and technical leads.
• Architecture stakeholders are a participant group of
those who want the architecture to successfully address
business needs, but who are not actively involved in the
evaluation process. This could include end users,
developers, and support staff.
In an ATAM, a software project is initiated when business drivers
identify a need for a system to address some problem. Business
drivers go hand in hand with the system architecture, which is
created as the solution to the business issues. Together, business
drivers and system architecture determine the quality attributes of
the system, the architectural approach taken, and the design
decisions that are made. These combine together to create quality
attribute scenarios.

Scenarios can then be analyzed, resulting in an evaluation of the
system, which includes trade-offs, sensitivity points, non-risk
scenarios, and risk scenarios. Since the risk scenarios have a
negative impact on the quality of the system, each of them are
analyzed and categorized into “risk themes.”
The diagram below illustrates this high-level process for ATAM.
The entire ATAM process itself can be broken down into nine steps.
1. Present the ATAM.
The evaluation team presents the ATAM process. This
includes: the context for the evaluation, expectations,
procedures, outputs, and addresses any concerns about the
evaluation.
2. Present the business drivers.
The project decision makers present the business problem
and the goals for the system as well as the system’s features
and requirements, project constraints, and scope.

3. Present the architecture.
Both current and expected state of the architecture is
presented as well as constraints such as time, cost, difficulty
of the problem, and quality expectations.
4. Identify the architectural approaches.
This analysis activity involves examining the architectural
patterns that have been used in the system. This step
analyzes the documentation and the notes from
presentations and asks questions to gain more clarity about
the system.
5. Create a quality attribute tree.
A quality attribute utility tree is created, which maps the
quality-related architecturally significant requirements
(ASR)s for each quality attribute. ASRs arise from the
business drivers.
To build such a tree, the overall “utility” of a system is broken
down into quality attributes, which are refined into attribute
refinements. Attribute refinements are more specific
qualities of a system. Once the quality attributes have been
refined, ASRs can be associated with the appropriate
attribute.
This step provides insight into the system and identifies
quality priorities. This step should be conducted in
conjunction with project decision makers.

In quality attribute trees, ASRs are given priority values to
denote if they are “must-haves” or not. The example above
uses high (H), medium (M), and low (L) designations, but
these values could differ from system to system.
6. Analyze the architectural approaches.
Using the prioritized ASRs from the utility tree, examine the
architecture and determine how it addresses each ASR. This
allows for the identification and documentation of risk and
non-risk scenarios, sensitivity points, and trade-offs. This
step reveals the capabilities of the system and consequences
of design decisions. It is not meant to be comprehensive, but
it identifies connections between business drivers and
system architectures.
7. Brainstorm and prioritize scenarios.
In this step, each group of participants creates quality
attribute scenarios that are important to them, which they
would expect when using the system. Scenarios with similar
quality concerns or behaviours can be merged.
This step provides an overview of the day-to-day system
usage and insights into the environment that the system can
be in.

Scenarios are prioritized based on importance to each
stakeholder, and the evaluation team compares the list with
the prioritized ASRs in the utility tree. If the priorities of the
stakeholders match closely with the priorities in the utility
tree, then there is good alignment.
There is a risk if there are many additional scenarios
discovered during this step that were not in the original set
of ASRs. It is an indication that the architecture is not able to
address the needs of stakeholders.
8. Re-analyze the architectural approaches.
Recreate a utility tree, but this time, using the top five to ten
scenarios prioritized in the previous step. This new tree can
be used to talk with the system architect and discover how
each scenario can be achieved with the system design.
9. Present the results.
Results of the evaluation are compiled together. This
includes all architecture documents, utility trees, risk and
non-risk scenarios, sensitivity points, trade-offs, and risk
themes. Risk scenarios should be grouped together by risk
theme. Risk themes help identify which business drivers are
affected.
ATAM helps expose unseen risks for stakeholders involved in the
architectural process.

Modern systems are becoming more and more complex, and
creating an architecture that can achieve all the requirements for
quality attributes is becoming increasingly important. Being able to
evaluate and analyze architectures helps successfully create high-
quality systems.
ATAM is a common method for analyzing and evaluating
architectures, especially as it does not require evaluators to have
intimate knowledge of the system, and covers the viewpoints of all
important stakeholders. ATAM helps minimize risks in a system by
identifying them and helps architects minimize the effects of
sensitivity points and be sensible about trade-offs.
ATAM also helps facilitate communication between stakeholders,
including identifying issues with newly discovered functionalities
that the stakeholders expressed to be important.
Relationship to Organizational Structure
A study from the Harvard Business School showed that looser
organizational structures, such as in open-source projects, led to
more loosely coupled modular code (MacCormack, Rusnak, &
Baldwin, 2007). On the other hand, in-house development teams
tended towards tighter coupling. They suggest that this difference
could be due to conscious choices by the developers or just a
natural consequence of the environment in which the software is
developed.
This study supports Conway’s Law. This law states that a software
system will tend to take a form that is congruous to the organization
that produced it.
This law should be taken into consideration when putting a software
development team together. Rather than bringing a team together
to work on a project with constant communication that can lead to a
tightly coupled system, knowledge of Conway’s Law should result in
planning development teams around the desired architecture.
Many programmers know Conway’s Law explicitly or intuitively.

When putting together a team of developers, it is a good idea to
first plan the architecture that works best for the system and
organize the team around that architecture. For example, if you are
building a web application using the n-Tier architecture, you could
have one team for the data backend, one team for the application
logic layer, and one for the presentation tier.
This increases the importance of flexibility in design, as there may
be cases where implementation of a component is not yet known or
may change.
Conway’s Law may entail extra work to provide unified and scalable
architectures. There may be need of a team whose purpose is to
consolidate and enforce common services across the whole system.
Even though Conway’s Law is called a “law,” it is actually
more of an observation that can be stretched or broken in
some situations. Even so, it is helpful to be aware of the law
and the implications it holds for putting together a team of
developers or software architecture.
Product Lines and Product Families
As developers work, they should constantly be looking for
opportunities for code re-use. Libraries, toolkits, and engines are
examples of how the software industry have made use of code re-
use. Code re-use, even if it requires modification, saves time and
resources, and it offers proven, robust solutions. This is why well-
written and well-documented code is important.
One means of harnessing code re-use is to treat a group of
products as a product line or product family. These terms are
somewhat interchangeable. This section will use product line.
Product lines are groupings of products that typically share the
same operating system, which allow for a great deal of code re-use.
A well-known example might be iOS, which is used across the
product line of iPhones, iPads, and iPod touch products.

Product lines present several advantages:
• Product lines designed with similar features allow companies
to save money because it reduces their development costs
per product. The time saved on development can be spent
testing for quality attributes, such as reliability, user
experience, security, and maintainability. However, less
testing is needed overall per product, because code that is
reused has already been tested.
• Another advantage of product lines is user experience. If
several products share characteristics, then the learning
curve for customers and developers is smaller. This may also
drive sales of other products in the same line.
• Product lines may also reduce time to market. Since most
software components are already made, making a new
product in the product line, or refining an existing one, takes
less time.
If only one or two products are being produced, it might not be
worthwhile to treat them as a product line, unless there is intent to
develop more products later on. In product lines, there is work
needed upfront to develop the foundations for products.
Implementing Product Lines
The book Software Product Lines in Action (2007) discusses
architectural considerations for software product lines. Let us
examine some of the steps outlined in this resource (see Course
Resources).
1. Commonalities
The features that stay the same in every product should be
identified.
2. Variations
The features of the product line that vary between products
should be identified. For example, for a line of tablets, a
variation could be support for different cellular network
connections.

3. Product specifics
The features that are unique to one and only product should
be identified. These features will be developed by the team
that is responsible for that product. For example, maybe one
of the tablets developed will also specialize in reading
eBooks. Its product specifics could be additional tools for
managing eBooks and support for an eInk section.
Below is a diagram of considerations for software product lines.

Once these considerations are established, you can turn to
development of product lines. Product line development teams are
generally divided into two camps:
• Domain engineering – This is the development of
commonalities and variations. Essentially, domain
engineering focuses on putting together the “infrastructure.”
• Application engineering – This is the actual development of
the product. It includes using the commonalities, deciding
which variations are necessary, and integrating them into the
product and developing product-specific features. This is
sometimes described as “instantiating” a product. There
could be several application engineering teams, one for
each product.
From a project management point of view, separating development
into domain and application engineering allows for separate
development cycles. The domain engineering team can determine
the long-term needs of the product line, evolve and test
components as needed, then release that infrastructure to the
application team. Then, the application team can develop product-
specific features using the infrastructure and the requirements of
the product. They will develop product-specific features, decide
what variations to use, and test the final product. This can be done
while the domain engineering team is working on the next update
of the infrastructure.

Reference Architecture
Products in a product line likely have similar architectures, as they
serve a similar purpose. Instead of creating a new architecture from
scratch for each product, the product line usually has an
architecture that products can build on or change. The domain
team is generally responsible for this reference architecture.

The reference architecture must:
• be designed with respect to the needs of the software, while
taking into account all current products in the product line
• include the capacity of variation, for differences in products
and to allow for future products
There are three general techniques to realize variations in a system:
• adaptation technique
• replacement technique
• extension technique
Adaptation technique. In this technique, a component has only
one implementation, but it supplies interfaces to change or add on
to it. Components can be adapted through settings in a
configuration file, by adding new methods or by overriding existing
methods. See below for a diagram of an adaptation technique.

Replacement technique. In this technique, there could be a default
component that is replaced with alternatives to realize variation.
There may not even be a default; instead there is a gap in the
system, and the developers must supply one of the components.
See below for a diagram of a replacement technique.
Extension technique. In this technique, a common interface is
provided for all the variations of the system. These variations are
often called extensions, add-ons, or plugins. See below for a
diagram of an extension technique.

Variations can be realized at different times. Different variations may
even come in at different stages—they can be formed early on,
during the design or development process of the application
engineering team, or they can be formed when the software is
compiled and built. Similarly, variations can be realized at different
times, like when the software is launched or after the software is
launched, whenever they are needed.
The dynamic nature of variations puts pressure on software
architecture to be highly modular. A combination of the nature of
the variation, the type of reference architecture, and the software
requirements determines what variation techniques are available to
architects, and how they are implemented. As with any design
decision, there are trade-offs to consider when building variation
into architecture.
This concludes the Software Architecture course notes. The next
course in this specialization will be focused on Service-Oriented
Architecture.

COURSE RESOURCES
Course Resources
• MacCormack, A., Rusnak, J., & Baldwin, C. (2007). Exploring
the duality between product and organizational architectures:
A test of the “mirroring” hypothesis. [Working paper 08-039].
Boston, MA: Harvard Business School. Retrieved from
http://www.hbs.edu/faculty/Publication%20Files/08-
039_1861e507-1dc1-4602-85b8-90d71559d85b.pdf
• Van der Linden, F. J., Schmid, K., & Rommes, E. (2007).
Software product lines in action: The best industrial practice
in product line engineering. Berlin, DE: Springer. [Books24x7
version]. Available from
http://common.books24x7.com/toc.aspx?bookid=31005

Glossary
Word Definition
Abstract data type A data type that is defined by the programmer and
not built into the language. It is a grouping of related
information that is denoted with a type, which allows
data to be organized in a meaningful way.
Abstraction A design principle that suggests that a concept in the
problem domain should be simplified down to its
essentials within some context.
Activities Actions that further the flow of execution in your
system.
Adaptation
technique
A technique to realize variations in a product line
system. In this technique, a component has only one
implement, but it supplies interfaces to change or add
on to it.
Application
engineering
The development of products in a product line.
Architecturally
significant
requirements
(ASR)
Requirements that arise from business drivers, which
are considered significant to the system architecture.
Architecture
stakeholders
A group of participants in ATAM who are stakeholders
that want the architecture to successfully address the
business needs but is not actively involved in the
evaluation process. This could include end users,
developers, and support staff.
Architecture Trade-
off Analysis
Method (ATAM)
A method for evaluating an architectural system that
can be used by outsiders of a system.
Artifacts A physical result of the development process.
Asynchronous A type of request-response relationship where a client
sends a request, but control returns right away, so it
can continue its processing on another need.
Attribute
refinements
More specific qualities of a system.

Availability The amount of time the system is operational over a
set period of time.
Ball connector Part of a component diagram that displays a provided
interface.
Behavioural design
patterns
Patterns that focus on how objects distribute work and
how they collaborate together to achieve a common
goal.
Client A program that requests a service or resource from
another program (the server).
Client/server
relationship
A relationship in which a client requests information or
actions, and the server responds.
Client-host The machine that hosts client software.
Clients A stakeholder of software architecture who supply
funding and outline the needs of the system.
Components Independent, encapsulated units within a system.
Conceptual
integrity
The concept of creating consistent software.
Conceptual integrity requires making decisions about
how a system will be designed and implemented, so it
seems as if only a single mind guided all the work.
Concrete scenario A concrete scenario is a quality attribute scenario that
is used to characterize a specific system.
Conway’s Law A law that states that a system will tend to take a form
that is congruous to the organization that produced it.
Creational design
patterns
Patterns that tackle how to create or clone new
objects.
Data accessor Any component that connects to a database that can
share data while operating independently,
communicate with the database through queries and
transactions, and save the new state of the system
back into the database using transactions.
Data-centric
software
architecture
A software architecture that allows data to be stored
and shared between multiple components, often
through the use of a database.

Data flow
architecture
A system architecture that considers the system as
simply a series of transformation on sets of data.
Data integrity The assurance that data is accurate and consistent
over its lifespan.
Data persistence The assurance that data will live on after a process has
been terminated.
Database A central repository for data.
Database
management
system (DBMS)
An automated system that manages and optimizes
queries and transactions.
Decomposition A major design principle of object-oriented modelling
and programming. It takes a whole thing and divides
it up into different parts. It could also take separate
parts with different functionalities and combine them
together to form a whole.
Designers A subgroup of the evaluation team. Includes those
involved in the architectural design of a system.
Domain
engineering
The development of commonalities and variations in
product lines. Essentially, the creation of the
“infrastructure” of a product line.
Encapsulation A fundamental design principle in object-oriented
modelling and programming. Encapsulation involves
three ideas: It bundles attribute values and behaviours
that manipulate those values into a self-contained
object; It also exposes certain data and functions of
the object to other objects; or alternately restricts
access to certain data and function to only within that
object.
End users A stakeholder of software architecture who will
ultimately use the software product.
Evaluation Team A group of participants in ATAM that includes
designers, peers, and outsiders.
Event-based
architectural style
A system architecture whose fundamental elements
are events.
Event bus The connector between event generators and
consumers in the system.

Event generators Also known as event consumers. Event generators
send events, and event consumers receive and
process these events.
Events Includes signals, user inputs, messages, or data from
other functions or programs. Events act both as
indicators of change in the system and as triggers to
functions.
Extension
technique
system. In this technique, a common interface is
provided for all the variations of the system. These are
often called extensions, add-ons, or plugins.
Feedback loop A situation where part of the output of a system is
used for new input.
Feedforward A situation where information from an upstream
process can be used to control a downstream
process.
Filters Entities in pipe and filter architecture that perform
transformations on data input they receive.
Flexibility Indicates how interchangeable other modules are for
a specific module. The more replaceable, the better
the design.
General scenario A general scenario is a quality attribute scenario that is
used to characterize any system.
Generalization A design principle that helps reduce the amount of
redundancy when solving problems, by taking
repeated, common, or shared characteristics between
two or more classes and factoring them out into
another class, so that code can be reused and
characteristics can be inherited by subclasses.
Good alignment When priorities of stakeholders in a software system
match closely with priorities identified in a utility tree
developed using the ATAM method.
Implicit invocation An invocation where functions are not in direct
communication with each other but where
communication is mediated by an event bus.
Instance level
diagrams
A deployment diagram that maps a specific artifact to
a specific deployment target.

Interoperability The ability of your system to understand
communications and share data with external systems.
Interpreter A component built into a system that can be used to
run scripts and macros and to drive programmable
actions specified by the user.
Kruchten’s 4+1
View Model
The combination of logical, process, development,
and physical views of a system along with scenarios
that provide an understanding of a system’s
architecture.
Latency The time it takes to produce an output after receiving
an input.
Logical view A view of a software system that focuses on
functionality of a system and the objects within it.
Maintainability Maintainability refers to the ease of modifying a
software system or component in order to correct or
improve faults, performance, or other attributes, or
the ability to adapt to a changed environment.
Manifestation A relationship where an artifact is a physical realization
of a software component.
Modifiability Modifiability refers to the ability of a system to cope
with changes to functions, incorporating new
functionality, or remove existing ones.
Nodes Deployment targets that contain artifacts available for
execution.
n-Tier architecture Also know as multitier architecture. It is a layered
architecture based on tiers.
Outsiders A subgroup of the evaluation team. Includes those
external to a project or organization.
Overhead A combination of computation time, memory,
bandwidth, processing, or other resources required to
perform a task.
Package A grouping of related elements in a software system,
such as data, classes, or user tasks.
Package diagram A UML diagram that shows packages and the
dependencies between them in a system.

Packageable
elements
Elements that can be packaged together, including
data, classes, or even other packages.
Peers A subgroup of the evaluation team. Includes those
who are in the project but not involved in design
decisions.
Performance The system’s throughput and latency in response to
user commands and events.
Physical view A view of a software system that focuses on the
physical components of software, and how they
interact and are deployed.
Pipe and filter
architecture
A type of data flow architecture that has filters that
perform transformations on data input they receive
and pipes that serve as connectors for the stream of
data being transformed.
Pipes Entities in pipe and filter architecture that serve as
connectors for the stream of data being transformed.
Principle of least
knowledge
An underlying design principle for design patterns. It
states that classes should know about and interact
with as few other classes as possible.
Process view A view of a software system that focuses on the
implementation and hierarchical structure of the
software.
Product line Also known as product family. A grouping of products
that usually share the same operating system, which
allow for code re-use.
Productivity Effectiveness of the effort of a team.
Project decision
makers
A group of participants in ATAM who are project
representatives with the authority to make project
decisions. This group could include project managers,
clients, product owners, software architects, and
technical leads.
Project managers A stakeholder of software architecture who plan,
identify, and mitigate risks, and manage the project.
Quality The degree of excellence of the software.

Quality attribute
scenarios
A scenario used to determine if a system is able to
meet the requirements that were set for the quality
attributes.
Quality attributes Measurable properties of a system used to gauge a
system’s design, run-time performance, and usability.
Reference
Architecture
An architecture for a product line that products can
build on or change so that each product does not
need to build an architecture from scratch.
Replacement
technique
system. In this technique, there could be a default
component that is replaced with alternatives to realize
variation. There may not even be a default; instead
there is a gap in the system, and the developers must
supply one of the components.
Request-Response The communication between a server and client,
where a client requests information or actions, and the
server responds.
Reusability Reusability refers to the ability to use existing
products and by-products of software development,
including code, designs, and documentation, more
than once in the software development process.
Sandboxing When different layers of a layered system run at
different levels of authorization or privilege.
Scenario Use cases or tasks required by end users of a software
system.
Script A description of the sequences of interactions
between objects and processes.
Security The system’s ability to protect sensitive data from
unauthorized and unauthenticated use.
Semaphore A variable or abstract data type that toggles between
two values: available and unavailable, which is used to
control access to shared data.
Sensitivity points Identifies processes in a system that could affect the
specific quality attributes of a system relative to an
ASR.

Separation of
concerns
A principle of software design, which suggests that
software should be organized so that different
concerns in the software are separated into different
sections and addressed effectively.
Server A program that provides a service or resource upon
request from another program (the client).
Server-host The machine that hosts server software.
Socket connector Part of a component diagram that displays a required
interface.
Software
Architecture
An aspect of the software development process,
which examines the high-level aspects of a system.
Software
developers
A stakeholder of software architecture who develops
and creates the software.
Specification level
diagrams
A deployment diagram that provides an overview of
artifacts and deployment targets, without referencing
specific details like machine names.
Stakeholders The people who have an interest in software, either
through use of the system or benefit from it.
Structural design
patterns
Patterns that describe how objects are connected to
each other.
Structured Query
Language (SQL)
A language used by relational databases that allow
queries to be asked of databases for information as
well as allow transactions to be performed.
Synchronous A type of request-response relationship, where a
client sends a request, then awaits the server’s
response before continuing execution.
Testability Testability refers to how easy it is to demonstrate
errors through executable tests.
Throughput The amount of output produced over a period of time.
Tier Components that are on different physical machines.
Often used interchangeably with layer, but this is not
correct.
Trade-offs An occurrence when one quality must be sacrificed or
restricted for improvement in another.

UML activity
diagram
A UML diagram that represents the control flow from
one activity to another in a software system.
UML component
diagram
A UML diagram concerned with the components of a
system.
UML deployment
diagrams
A UML diagram used to visualize deployment details
of a software system.
Usability The ease at which the system’s functions can be
learned and used by the end users.

Software-Architecture_Course-Notes.pdf

Recommended

More Related Content

Similar to Software-Architecture_Course-Notes.pdf (20)

Recently uploaded (20)

Software-Architecture_Course-Notes.pdf