Notes taken from:
Object Oriented Software Engineering text;
Art of Software Architecture (Stephen Albin) Use Case Analysis (Bittner and Spence);
Use Cases – Requirements in Context (Kulak and Guiney) Some notes from Rational Software Corporation slides
Non-Functional Requirements
Supplement the Use-Case Descriptions
For each use-case realization
Find Classes from Use-Case Behavior
Distribute Use-Case Behavior to Classes
For each resulting analysis class
Describe Responsibilities
Describe Attributes and Associations
Non-Functional Requirements – We have a:
good understanding of the nature of the analysis classes and their responsibilities
collaborations required to support the
functionality described in the Use Cases via analysis classes
Need to address the non-functional requirements
Use-Case Analysis Steps – Here’s
where we are:
Non-Functional Requirements
The purpose of “identifying non-functional requirements” is to get a handle on these absolutely necessary requirements that are normally not ‘functional.’
During Analysis, this information is speculative.
Will be refined later.
Philosophy: Capture now; Realize later…
(Essential points in design…)
Non-Functional Requirements –
OOSE Text p.115-
Types of Requirements
Functional requirements
Describe what the system should do
Non-functional requirements
Consists of Constraints that must be adhered to during development (design and
implementation)
Operative word: ‘Constraints.’
Equivalently: these are design constraints
Functional requirements
What inputs the system should accept
What outputs the system should produce
What data the system should store that other systems might use
What computations the system should perform
The timing and synchronization of the above
Non-Functional requirements
Mostly not unique to a single use case.
Generally thread themselves throughout some use cases
Transcend locality
Some of these conflict with each other
Some support ‘efficiency’ of code and
‘inefficiency’ in maintenance…
Let’s contrast functional and non-functional
requirements
Non-functional Requirements – Verifiable; Measured
All requirements must be verifiable
If not ‘verifiable’ then there are no indications that these requirements have been satisfied.
Some must also be measured.
Some requirements are measurable, such as response time;
availability time, etc.
Some may be directly measured; some measured via simulation.
We typically parse the non-functional requirements into categories so we can address them.
Categories of Non-Functional Requirements
1. Quality Non-Functional Requirements (1 of 2)
All constrain the design to meet desired quality levels of usability, efficiency, reliability, maintainability and reusability…
Examples:
Response time – particularly important for processes that process a lot of data or use networks a great deal.
Requirements might stipulate < two second response.
Might use a Timing bar indicating progress…
Response time might be considered a functional requirement for some ‘real time systems.
Throughput – for computationally-intensive applications or for heavy transaction-oriented systems,
the ‘number of transactions per minute or number of computations per minute can/must be measured.
Resource usage –
Application cannot use > xxMB of memory
Very practical in a large multiprogramming / multiprocessing system.
Categories of Non-Functional Requirements 1. Quality Non-Functional Requirements (2 of
2)
Reliability – typically Mean Time Between Failures (MTBF) (number of failures per week; define ‘failure.’)
Availability – Measures the time application is ‘up’ and the period of down time, when down.
Recovery from failure – Mean Time To Repair: (MTTR) – critical to some applications; not so for others; checkpoint restarts; auto-saves, audit trails, etc. If you specify the
detailed procedure to follow, then this is a functional requirement.
Allowances for Reusability – may specify that a certain percentage to be designed generically to support
reusability…
Categories of Non-Functional Requirements 2. Environment and Technology Constraints
Platform
What hardware / software the software must be able to work on.
Normally indicate ‘least’ platform
e.g. Windows 7, Vista; 120 MB free disk space….
Technology to be used
Programming language; database system. All individuals are familiar with these languages / database nuances.
May/may not reduce need to train people
Often, these are specified in a work environment.
Categories of Non-Functional Requirements 3. Project Plan and Development Methods
Development Process/Methodology
This may be specified and will greatly impact development.
Waterfall; RUP; Agile Methods, Scrum; XP, Feature-Driven Development (FDD), etc.
Particular approaches to testing (done by developer?
separate group; in-house? External?
References to these must be included; not the details
Cost and Delivery Dates; Traceability
A Bigee!! Usually a very important constraint.
Usually found in the Contract for the system or in a
separate project plan.
Random Thoughts - Non-functional Requirements
Not always clear that a requirement is non-functional and functional.
If requirement is a fundamental part of the application’s functionality, then it should be stated as a functional
requirement
Think ‘performance’ of a real time / manufacturing app.
Think ‘reliability’ of a safety-critical application
If requirement is a ‘constraint’ on design or some kind of restriction on design, then the requirement is almost
always ‘non-functional.’
Certainly the presented list is not exhaustive!
Quality Models – Many and Varied (1 of 2)
Quality Models from a different source:
from Art of Software Architecture…
(p.70-72)
Different ‘categories’ for Quality Modeling according to this source:
Product Operation Quality:
Accuracy, Reliability, Efficiency, Integrity, and Usability
Product Revision Quality are:
Maintainability, Flexibility, Testability
Product Transition Quality are:
More attributes…(2 of 2)
According to Barry Boehm: Overall quality is a function of these metrics:
Validity, clarity, understandability, modifiability, modularity, generality, economy, resilience, documentation
Laurence Best’s Application Architecture identifies:
Accuracy and comprehensiveness; simplicity of use,
operational flexibility, ease of development, maintenance and extension, security and control, resource efficiency, recovery
And there are numerous other ‘lists’…..
These are presented to show you that there are different
‘takes’ on quality and what determines quality…
Looking down on these…(1 of 2)
Functional models such as use cases, typically drive object-oriented projects (although they certainly don’t have to…)
Attributes such as modifiability are not easily expressed in purely object-oriented terms and need supplemental text descriptions to represent the design.
Each attribute MUST be considered during design and usually requires the architect to make multiple passes as a design.
Class design: start by addressing functionality only
Second pass to incorporate modifiability constraints while making sure that the functionality is not compromised.
Sometimes, multiple passes are needed to incorporate various requirements and to evaluate the design tradeoffs.
Looking down on these…(2 of 2)
How can these methods be efficient and yet maintainable?
Efficient and yet reusable?
Reusable yet modifiable
Many competing non-functional requirements.
Architect’s design must realize a balance among them.
Example:
Performance design techniques, however, usually incorporate fewer levels of indirection and optimizations to data structures in order to improve operation execution times.
Performance gains may bring decreased maintainability / reusability features.
Cost is usually competing with everything else and it is common
Where are Non-Functional
Requirements Captured / Modeled?
Use Case Modeling – Bittner and Spence – p. 43
In the UP, we place the non-functional requirements in the In the UP Supplementary Specifications
But this is usually considered a ‘standard’…..
Consider: (Bittner and Spence: still different categories!!!)
Legal and Regulatory Requirements
e.g.: The customer must be of legal age to purchase alcohol.
Application Development Standards
e.g.: The system must be developed in accordance with the Rational Unified Process
Quality attributes of the system to be built must include usability, reliability, performance, supportability requirements
e.g.: The system must be available at least 90% of the time.
Constraints placed on the design and implementation of the system such as operating system, environments, …
Supplementary Specifications
Inferior or Lesser.
Very misleading to think ‘supplementary specs’ are inferior or lesser than functional specs.
Amuck projects.
Many projects have run amuck because they did not consider these requirements.
Major part of success criteria.
May have formed a major part of the overall critical success criteria.
Supplementary Specifications
Consider the following true story:
System must support 300-500 simultaneous users;
On entry into user acceptance testing, system supported eight simultaneous users;
Developers had concentrated on the user interface and completely ignored this supplementary spec item.
Upon further review, it was noticed that the
developer did not investigate any of the requirements beyond the scope of the most basic flow of events.
“Not surprisingly, system was never deployed.”
Supplementary Specifications – Complement to Use Cases
Note use cases can capture non-functional requirements especially those that apply within the context of a single use case.
Use Supplementary Specs to capture non-functional
requirements that are global in nature or do not vary from one use case to the next.
Keep a balance between using the use cases and non- functional requirements.
We need both if the overall requirements objectives are to be met.
Non-Functional Requirements – the Process
Use Cases – Kulak and Guiney…
Identify the non-functional requirements
Validate with appropriate stakeholders
Document the requirements
Capture its name, its effect, both the system and business
degradation that would result if the requirements are not satisfied
Best time to capture them when exploring the functional requirements
Document it right away from your sources of information (interviews, questionnaires, …)
Ask users about response times, what bothers them now; what is fine.
Be certain to differentiate between business-related concerns which can often be added to a use case description, with non- functional requirements.
Non-Functional Requirements – the Process
Rank the non-functional requirements. Very Important!
More Expenses than worth. Sometimes
accommodating non-functional requirement may cost more than implementing the use case.
No free lunch. Some things are not worth the cost to accommodate.
e.g. Infinite availability; levels of performance. “Don’t let [a user’s] initial assumptions drive you into an architectural nightmare.”
Use a template to document the non-functional
requirements, as found in our Use Case book, ~ p. 79.
FunctionalNon- Category*
Number Non Functional
Requirement Name Description Applies to Use
Case(s): Exceptions
Use something like the table below.
You may add ‘column attributes’ as needed.
Non- Functional
Category*
Number Non Functional
Requirement Name Description Applies to Use
Cases: Exceptions
1 Persistency Student objects need
to be persistent Where Student Objects are
Updated
None
2 Security System Administrator
actor needs to be authenticated
Those Use Cases that access the Profile Database
None
3 Availability The Profile Database must be available 95%
of the time
All Use Cases that require authentication
None
Examples
Supplementary Specifications
Inserted table(s) into Supplementary Specifications.
The Supplementary Specifications document is an additional artifact that we will be using.
Note also that in some cases, the non-functional requirements are called ‘analysis mechanisms’
(RUP)
Later, we will talk about; ‘persistency mechanism’;
security ‘mechanism,’ etc. Means that there will be some process to accommodate persistency, security, etc.
Analysis Classes – Use of tags { … } Non-Functional Requirements Map UML –
•
As analysis classes are identified, it is important to identify the non-functional requirements that might apply to them• Classes that must be persistent are mapped to the Persistency Mechanism and tagged as persistent. Persistency tag: {persistent}
• Classes that are maintained with the Legacy Course Catalog system are tagged to the Legacy Interface mechanism;
• Classes for which access must be controlled (like who can read and modify instances of the class) are mapped to a Security
mechanism., etc. {security} or {secure mechanism}
• Classes that are distributed classes mapped to a Distribution
Supplement the Use-Case Descriptions
For each use case realization
For each Analysis Class describe
responsibilities, etc….
…..
Know (now) about analysis classes
their responsibilities,
the analysis mechanisms / non-functional requirements that these analysis classes need to implement (persistence, security, legacy,
…) and
the collaborations required to support the functionality described in the use cases.
Unify Analysis Classes: Now lets review everything to ensure it is complete and consistent before moving on….
Checkpoints
Use-Case Analysis Steps
<<control>>
<<boundary>>
<<entity>>
<<entity>>
Unify Analysis Classes
The purpose of “Unify Analysis Classes” activity is to ensure that each analysis class represents a single well-defined concept, with non-overlapping responsibilities.
Name of analysis class should capture role; (e.g. EnrollmentForm)
Description of class should capture role played by class in the system.
Merge classes that define similar behavior or represent the same thing.
Merge entity classes that define the same attributes
Aggregate behaviors of merged classes.
If you do any of these things, make sure you update any supplemental use case descriptions where
Supplementary Specification
Glossary
Analysis Classes
Evaluate Your Results
Are the Names appropriate?
Are the Descriptions accurate?
Are classes that should be merged actually merged?
Evaluating and Verifying
Now, verify analysis classes meet functional requirements of the system.
Trace down through the Use Case specifications for each Use Case.
Are all responsibilities covered somewhere? Are domain / application objects captured? Properties?
Verify the analysis classes and their relationships are consistent with collaborations that they may support.
If one class is sending messages to another, there is a collaboration needed. If one class defines an ‘array’ of objects from another class, have we modeled this as an aggregation!
Very important that you evaluate your results at the conclusion of Use Case Analysis.
Use-Case Analysis Steps
Supplement the Use-Case Descriptions
For each use-case realization
Find Classes from Use-Case Behavior
Distribute Use-Case Behavior to Classes
For each resulting analysis class
Describe Responsibilities
Describe Attributes and Associations
Qualify Analysis Mechanisms
Unify Analysis Classes
Checkpoints - Check the ‘quality’ of the model
against criteria that the Designer looks for…
Checkpoints: Analysis Classes
Are the classes reasonable?
Does the name of each class clearly reflect the role it plays?
Does the class represent a single well-defined abstraction?
Are all attributes and responsibilities
functionally coupled? (What does this mean to you?)
Does the class offer the required behavior?
Are all specific requirements on the class
addressed?
Checkpoints: Analysis Classes
Note: All checkpoints should be evaluated with regards to the use cases being developed for the current iteration.
The class should represent a single well-defined abstraction. If not, consider splitting it.
The class should not define any attributes and responsibilities that are not functionally coupled to the other attributes or responsibilities defined by that class.
The classes should offer the behavior the use-case realizations and other classes require. (These will be methods later…)
The class should address all specific requirements on the class
from the requirement specification – the Use Cases in our case.
Checkpoints: Use-Case Realizations
Have all the main and/or sub-flows been handled, including exceptional cases?\
Have all the required objects been found?
Have all behaviors been unambiguously distributed to the participating objects?
Have behaviors been distributed to the right objects?
Where there are several interaction diagrams, are
their relationships clear and consistent? (next
Review: Use-Case Analysis
What is the purpose of Use-Case Analysis?
What is an analysis class? Name and describe the three analysis stereotypes.
What is a use-case realization?
Describe some considerations when allocating responsibilities to analysis classes.
Describe the role of non-functional requirements in complementing functional requirements.
