Being a very brief history of how "architecture" become a thing in software, and of how it delivers on its core claim to fame, which is:
Enabling you to Reason & Calculate about quite vague "Quality" requirements and thereby to achieve confidence of a successful system and happy customers
2. In the beginning ... ...
there
were
algorithms;
computable
functions;
the lambda
calculus…
3. The Software Architectural
Qualities of a Proof on
Paper…
• Availability: 7–Nines Uptime
• Reliability/Robustness: no
moving parts, can survive drops,
collisions and earthquakes
• Parallel processing: Can be
copied and worked on by
multiple people in parallel (good
implementation of immutability)
• Can be geographically
distributed without loss of
performance
• But not, alas, extremely fast…
4.
5.
6.
7. why architecture? because …
“… No-one replaces or re-writes a system
because of its functionally. It’s always because
of some quality failing – performance or
reliability, usability, or ease of modifiability”
8. The Claim of Software Architecture
To enable Reasoning about:
… the Quality Attributes of
Software-intensive Systems
… to meet Stakeholder’s Concerns
… using Architecture Descriptions
of the system’s fundamental
structure(s)
in terms of its elements, relationships, properties and principles
Why software architecture?
9. • What is a Quality Attribute?
• What does “Reasoning about” mean?
• What are Stakeholder Concerns?
• What is an Architecture Description
and how does it help?
4 Key Concepts The promise of Software Architecture
10. • What is a Quality Attribute?
• What does “Reasoning about” mean?
• What are Stakeholder Concerns?
• What is an Architecture Description
and how does it help?
4 Key Concepts The promise of Software Architecture
11. What is a Quality Attribute? Who defines quality?
“It’s not what you do, it’s the way that you do it”
affordability, availability, correctness,
deployability,efficiency, evolvability, extensibility,
fault-tolerance, main-tainability, modifiability,
reliability, resilience, responsiveness, robust-ness,
safety, scalability, securability, testability, usability,
…
12. What is a Quality Attribute? ISO 25010
“It’s not what you do, it’s the way that you do it”
accessibility, accountability, accuracy, adaptability, administrability, affordability, agility, auditability, autonomy,
availability, compatibility, composability, configurability, correctness, credibility, customizability, debugability,
degradability, determinability, demonstrability, dependability, deployability, discoverability, distributability,
durability, effectiveness, efficiency, evolvability, extensibility, failure transparency, fault-tolerance, fidelity, flexibility,
inspectability, installability, integrity, interchangeability, interoperability, learnability, maintainability, manageability,
mobility, modifiability, modularity, operability, orthogonality, portability, precision, predictability, process
capabilities, producibility, provability, recoverability, relevance, reliability, repeatability, reproducibility, resilience,
responsiveness, reusability, robustness, safety, scalability, seamlessness, self-sustainability, serviceability,
supportability, securability, simplicity, stability, standards compliance, survivability, sustainability, tailorability,
testability, timeliness, traceability, ubiquity, understandability, upgradability, usability
13. Many ways to get from A to B what about a horse?
15. What is a Quality Attribute? but … what do all these words mean?
“It’s not what you do, it’s the way that you do it”
accessibility, accountability, accuracy, adaptability, administrability, affordability,
agility, auditability, autonomy, availability, compatibility, composability,
configurability, correctness, credibility, customizability, debugability, degradability,
determinability, demonstrability, dependability, deployability, discoverability,
distributability, durability, effectiveness, efficiency, evolvability, extensibility, failure
transparency, fault-tolerance, fidelity, flexibility, inspectability, installability, integrity,
interchangeability, interoperability, learnability, maintainability, manageability,
mobility, modifiability, modularity, operability, orthogonality, portability, precision,
predictability, process capabilities, producibility, provability, recoverability, relevance,
reliability, repeatability, reproducibility, resilience, responsiveness, reusability,
robustness, safety, scalability, seamlessness, self-sustainability, serviceability,
supportability, securability, simplicity, stability, standards compliance, survivability,
sustainability, tailorability, testability, timeliness, traceability, ubiquity,
understandability, upgradability, usability
16. Software Quality Attributes first, define your terms
Define what matters to you about a quality attribute
by either
Describing scenarios or
Defining measures (or proxy measures)
Then research the tactics/patterns/appliances you
can use to achieve the quality
17. Availability & Resilience
Availability / Resilience
o A common measure is up-time: “5-nines”
o A scenario for a failure event might be
“When an out-of-memory failure occurs, the app should
recycle, and the system should continue operating as
normal within 30 seconds.”
o Attack Scenarios:
“When attack X happens, Y should happen and the
system should continue operating as normal”
quality attributes : define it
18. Modifiability is ... quality attributes : measure it
Modifiability / Maintainability / Evolution
o Could be measured as (estimated) cost or speed
Are these reasonable measures of modifiability?
o “It should not require a change request and 14 day
wait to correct a spelling error on the website”
o “Sprint velocity after a year of development should be
at least as good as in months 3 to 12”
19. • What is a Quality Attribute?
• What does “Reasoning about…”
mean?
• What are Stakeholder Concerns?
• What is an Architecture Description and
how does it help?
4 Key Concepts The promise of Software Architecture
20. and even to predict
What is “Reasoning
about…”?
estimate
measure
risk-evaluate
account for
cost-benefit-analyse
calculate
quantify
validate
budget
everything
21. What is “Reasoning about…”? Show me the numbers
♣ Being able to describe, define,
measure, calculate, test and even
predict how a system will behave …
♣ … preferably in advance of paying
the full cost of delivery and preferably
before missing the timeframe in which
the system is still useful
24. • What is a Quality Attribute?
• What does “Reasoning about…” mean?
• What are Stakeholder Concerns?
• What is an Architecture Description and
how does it help?
4 Key Concepts The promise of Software Architecture
25. 4 Key Concepts Such many stakeholders
Not only the customer & the end-user but also…
28. 4 Key Concepts The promise of Software Architecture
• What is a Quality Attribute?
• What does “Reasoning about…”
mean?
• What are Stakeholder Concerns?
• What is an Architecture Description
and how does it help?
41. 20 years on … “6 + 0 + 1”
Rozanski & Woods, Software Systems Architecture, 2nd ed
42. The Claim of Software Architecture
To enable Reasoning about:
… the Quality Attributes of
Software-intensive Systems
… to meet Stakeholder’s Concerns
… using Architecture Descriptions
of the system’s fundamental
structure(s)
in terms of its elements, relationships, properties and principles
Why software architecture?
43. What is “The Architecture” of a
system?
rough cut definitions
“… the fundamental structures or
organisation of your code”
“… all the rules & design decisions you have
get right up-front, because they are too
expensive to change later.”
44. Architecture is ... Bass, Clements, Kazman, 1997-2012
SEI, early 2000s
“The structure or structures of the system,
which comprise software elements, the
externally visible properties of those elements,
and the relationships among them.”
45. Architecture is ... what does fundamental mean?
ISO/IEC/IEEE 42010:2011
“The fundamental concepts or properties of a
system in its environment embodied in its
elements, relationships, and in the principles
of its design and evolution”
46. Architecture is ... Kruchten, updated 2009
Kruchten 2009: The significant decisions about
♣ the organization of a software system,
♣ the selection of the structural elements and their
interfaces by which the system is composed together with
their behavio[u]r as specified in the collaboration among
those elements,
♣ the composition of these elements into progressively
larger subsystems,
the architectural style that guides this organization, these
elements and their interfaces, their collaborations, and their
composition
47. How to produce an architecture A catalogue & a method
How can we design systems to meet quality
requirements?
1) A catalogue of off-the-shelf tactics,
components, appliances, patterns,
frameworks, reference architectures… with
known quality–impacts
2) A method for applying them when
designing your system
48. catalogues of patterns, tactics, …
A (very long) Reading List
http://bing.com/search?q=architecture+tactics
http://bing.com/search?q=reference+architectures
http://google.com/search?q=architecture+pattern
http://pubs.opengroup.org/architecture/togaf8-doc/arch/chap28.html
Rozanski & Woods, Software Systems Architecture
Bass et al, Software Architecture in Practise
Buschmann et al,
Pattern-Oriented Software Architecture Vols 1–5
Blue Peter Architecture
49. Patterns,Tactics, Appliances Off the shelf architecture
Including …
Pipe and Filter Architecture, Blackboard Architecture, Hexagonal Architecture,
Layered Architecture, Service Oriented Architecture, Message Oriented
Distributed Architecture, Client-Server, Client-Proxy Server, Master-Slave,
Microservices, Reactor, Replication, Transactions, ACID, Eventual Consistency,
Subsystem Interface, RDBMS, NoSQL DB, Graph Database, Datawarehouse,
GUI, Command Line Interface, Batch Processing, Thread Pool, Activator, Factory,
Dependency Injection, File System, Web Application, Sharding, Horizontal
Scaling, Load Balancer, Message Queues, Publish-Subscribe, Request Broker,
Web Services, Desktop Applications, Mobile Apps, Virtual Machine, Model View
Controller, Domain Model, Domain Driven Design, Immutability, Caching, Proxy,
Facade, Bridge, Adapter, Mediator, Decorator, Command, Interpreter, Web Farm,
Web Garden, P2P, ORM, Data Mapper, Active Record, Tokens, Locks, Monitors,
Heartbeat, Error detection, Backup and DR, the InterWebs, LAN, WAN, AES, PKI,
Flash Drives, Gigabit Ethernet, WPA2-PSK
50. a method for designing an
architecture
1. Draw a box for the system. Surround it
with the external actors which/who must
interact with it.
2. List the Architecturally Significant
Requirements in priority order
3. Start at the top of the list and work down
4. For each requirement, Identify or Create a
Partition of the system which will meet
that quality requirement
5. Identify tactics/patterns/appliances which
achieve the Quality Requirement
6. Goto 2
Simplified from Bass et. al.
51. Simple Case Study No relation to Uber
Our New Startup
• sÜper is an application which will find other
sÜper users near you, and help you to
arrange a lift share.
• Business plan: Grow users to critical mass in
year 1, charge a fee from year 2 onwards
52. Functional & Quality Requirements
Key Functionality
•Based on the sUper user’s geolocation, find their nearest peers
•Enable communication & appointment making between
Quality Requirements
• Must be mobile/geo-based
• Must be very very simple & fast to use & not make
embarrassing mistakes
• Must be virally attractive
• Must (appear to) be reliable
• Available 24/7
• Must be cheap to build & maintain for the 1st year
• Must cope with viral growth in user base
• Must keep user data especially payment data secure
Are not un-related to each other
53. ASRs
Prioritised Architecturally Significant Requirements
1. Deployability: Must run on mobiles
2. Usability: Must be very very simple
3. Usability/Performance: Must be fast to use
4. Usability/Accuracy: Not make embarrassing mistakes
5. Low Initial Cost & Time to Market
6. Functionality: Communication between users
7. Functionality: Find nearby users
8. Scalability
9. Availability 24/7
because the architect said so …
“The Beginning” for computer science was the 1930s: the attempts to formalise mathematics resulted in Church’s Lambda Calculus and Turing’s Machine. All of which was done on paper…
Computer was a real person’s job title until the 1940s. The Manhattan project on the atomic bomb had teams of people, and large mechanical calculators
What it doesn’t have is speed….
Bletchley Park needed speed for breaking first the Navel Enigma, and then the Lorentz cipher. They built first ‘bombes’ – racks of automated enigma machines running in parallel – and then:
The world’s first electronic programmable digital computer
Machines are like Regular Expressions. As in, “Some people, faced with a problem, think, ‘ah, I’ll use a regular expression’. Now they have two problems”.
If you have a problem and you build a machine to solve it, you now have two problems. Or more.
How do I setup or program the machine? How much does it breakdown? Who can repair it? How often does it need maintenance? What does that cost? Is it fast enough? Will it overheat? Or explode? What does this flashing light and smell of burning moth mean?
The first bug was hardware.
A moth that was caught in a mechanical relay of the US Navy Mark II, 1945
But with software systems, you have 3 problems:
Things that used to go wrong in machine… now go wrong in software as well
Functionality is what you do: calculate the square root of an integer; Do this month’s payroll and pay everyone and pay the taxman. Quality is how you do it.
Functionality is what you do: calculate the square root of an integer; Do this month’s payroll and pay everyone and pay the taxman. Quality is how you do it.
There is an unlimited number of ways to meet a given functional requirement
If your requirement is to get from A to B as fast as possible … with 50 friends … then the bus is a better solution. The one seater car is less fun if you have to wait 10 hours in the rain for your turn to drive.
But.
You want your system to be scalable? What does that mean?
Every one of these words is so vague and ill-defined as to be completely useless.
Examples of a measure and a scenario
In software architecture, as in politics, the mark of the expert is firstly having all the numbers at your fingertips and secondly being able to communicate them in a compelling and easy-to-grasp way.
Show me the numbers. This is good reflex reaction to have in many areas of life, especially politics.
Abstraction is the key to reasoning and quantifying. The trick is to ignore almost everything, except just the one thing you care about.
If we want to get 50 people from A to B we can (almost) completely ignore the tyres, the engine size, the brakes, the electrical systems, the 0-60 time and focus on just one thing: N=number of seats. This is the power of mathematical abstraction; by ignoring 99% of what’s in front of you and focussing on one thing you can “prove” that the bus is the better solution.
Ignore everything about wheels, tyres, axles, chassis, engine, power to weight ratio, fuel, and focus on just one thing about each vehicle: how many seats?
Architects talk about stakeholders.
This expands on having just a customer, who is the main stakeholder mentioned in the agile manifesto; and includes end-users, other areas of business with legitimate concerns such as the information security officer, the people on help desk, the application support developer 2 years later, yourselves as the development team, the CFO and so on.
Architecture claims in general to be able to address or help to address many of their concerns by considering stuff considered architecturally significant.
I was once a café manager for 6 months. I was possibly appointed to this rôle because of my enthusiasm for coffee. It would all have been so much easier if it weren’t for the customers.
Can you build a car, and then put go faster things on it to make it go faster? Or do you have to address gofasterness right from the start?
If you have a set of user stories on a Monday morning as you consider the sprint plan, are they all equally risky or do some stand out with a little flashing light in your head that says ooh that’s a problem?
So, a kind of triage: identifying stuff you have to deal with sooner otherwise you won’t be able to deal with it at all.
Continuous delivery is an example that agile has highlighted: If you don’t get your CI pipeline going early on, it gets harder and harder and costs you more and more.
I go back again to the bus vs “racing” car example, of abstracting the number of seats
Ignore everything about wheels, tyres, axles, chassis, engine, power to weight ratio, fuel, and focus on just one thing about each vehicle: how many seats?
Or then again, if you only want to get 1 of you from A to B as fast as possible
You might want just two numbers: power and weight
An Architecture Description must use as many abstractions / aspects / structures as the significant quality attributes require. We call these ‘aspects’ or ways of looking at the system, viewpoints
There is no single, all-encompassing viewpoint
If you want some assurance that supporting a new bitmap graphic format will not require rewriting half the system, then this picture would give you considerable comfort.
It suggests that you would only have to add one teeny weeny box and a single dashed line, and you’re done; with no alternations to any other part of the system.
“If we modularise the code well and apply encapsulation then in a year’s time we hope to be able to add or modify functionality as quickly/easily/reliably as we could when it was a greenfield project.”
But if you’re concerned about taking a website from 100 users per day to a 1,000 per day and more, then this diagram is much more important to you
“If we design the app to run on a webfarm behind a load balancer, we can scale up to millions of users”
Different views of the system give you assurance about different quality attributes
One of the most popular sets of Viewpoints is 4+1 (google it)
And finally …
What is An Architecture? And how do I design one?
The bit that isn’t on mobile will probably have to communicate with all the mobiles, so will probably end up as a server of some kind
Options
Choose a high quality cross-platform mobile UI framework => Trial Xamarin & PhoneGap
- Choose ‘best of breed’ for each phone platform (Which is what?)
Domain Model will impact usability
There are no magic tricks for making phones go faster, but may have to prefer native apps not html/javascript?
Minimize what has to run on the phone?
What technology are we most familiar with?
If Javascript, then use PhoneGap+Node ?
If .Net, then use Xamarin+ asp.Net ?
- Source geolocation & map libraries
- Trial Google maps and/or bing maps
Must be something off-the-shelf that helps with this. We probably want bi-directional comms.
WebSockets? Some kind of Message Queue?
4+7 look like a domain model?
Load balance the server.
We shouldn’t need to/can’t scale anything client side BUT we must ensure that comms does not grow exponentially with number of users!