Software Architecture

Planning a system before it is built

High-level overview of components

Christopher Alexander

A Quality
Without a Name

Feeling of satisfaction and contentment engendered by good building

Design patterns

Gang of Four Patterns

• Solutions to common problems
• Specific small-scale scenarios
• Workarounds for Java limitations

Architectural style

How you approach the problem?

Form follows function ideal of modernism

Architectural patterns

Thinking about large software systems

How is system divided into components?

How is development divided into teams?

How components communicate with each other?

Styles do not prescribe specific implementation!

Component architectures

What are the components & how they communicate

How is the problem divided into component?

What components talk to what other components?

How is the communication set-up?

What teams work on what parts of the system?

Layered architecture

Components in several layers

Can only call components from one layer below!

Layered architectures

Typical use cases for the pattern

Operating system with kernel, system, user processes

Enterprise with presentation, business & persistence

Web apps with web page, server and database

Network stack TCP/IP protocol and such

Layered architecture

Potential advantages

• Can replace individual layers
• Layers need only limited info
• Can be built independently

Potential problems

• May lead to inefficient systems
• Changes may be hard to make

Three-tier architecture

Client-side code
Server-side code
Database

Communication over network

Different langua- ges in each layer

Service-oriented architecture

Services that call each other

Different services implement different functionality

Service-oriented architectures

Services in a sample retail system

Order management service to take orders

Warehouse management service to manage stock

Shipping service to handle delivery

Customer relationship service for communication

Service-oriented architectures

Implementing SOA

• Service description
• Service discovery
• Service deployment

Pros and cons

• Separation of logic and teams
• Can get heavy-weight and complex

Microservices

A more modern take on service-orientation

Simpler communication via REST

Each service can use different language

Still need discovery and message queues

Can use auto-scaling in cloud (Kubernetes etc.)

Publish subscribe pattern

Information exchange

• Who needs to be called?
• Publish events to anyone
• Subscribe by category or type

Application areas

• Information exchange in finance systems
• RSS feeds for blogs & news
• Anything aiming for large extensibility

Information management

How can information be managed & stored?

How to pass information in complex processing?

How to facilitate efficient sharing?

What component can access what information when?

How to avoid losing information on error?

Pipeline architecture

Data processed in a sequence of steps

Think UNIX pipelines!

Data science and scientific computing

Pipeline architecture

Potential benefits

• Small composable blocks
• Different ways of connecting
• Easy to add more tools

Open questions

• Format of data passed in between
• UNIX shell vs. Microsoft PowerShell
• Works if data is of a same kind

Blackboard architecture

Data shared globally

Many different processors can access and transform it

For example, analysing natural language

Pipeline architecture

Implementation

• Choosing common data representation
• Resolving conflicting solutions
• Component implementation

Application areas

• There is no complete solution
• Natural language processing
• Classic AI like expert systems

Event sourcing architecture

Any action in the application is an event

Store events,
not final state!

Banking systems that cannot lose data

Event sourcing architecture

Implementation

• Suitable event representation
• Event store to keep events
• State * Event -> State function

Application areas

• Financial software systems
• Functional programming (fold)
• Reactive programming (Elm language)

Large software architectures

Software architecture and patterns
Seeking quality without a name
Gang of Four patterns and beyond

Component-based architectures
Layers and three-tier web applications
From services to microservices

Information-oriented architectures
Pipeline in UNIX and Blackboard in AI
Event-sourcing architecture for reliability

CO559: Large software architectures

What you should remember from this lecture

• Different architectural patterns discussed
• Microservice are widely used today!
• Event sourcing is great in certain areas!

Tomas Petricek
t.petricek@kent.ac.uk | @tomaspetricek

Books

• Richard P. Gabriel (1996) Patterns of Software
• Frank Buschmann et al. (1996) Patterns-Oriented Software Architecture, vol 1