Skip to content

Software Architecture

Overview of Software Architecture

  • Definition: Software architecture is fundamental to business survival, reliability, scalability, and maintainability. It guides the long-term evolution of a system.
  • Levels of Abstraction:
    • Small-level: The architecture of individual programs.
    • Large-level: The architecture of complex enterprise systems.
  • Advantages of Design and Documentation:
    • Facilitates stakeholder communication.
    • Enables effective system analysis.
    • Supports large-scale reuse of components and patterns.
  • Modelling Techniques: Architectures are often modelled using simple block diagrams, UML, or other detailed diagram representations.

Case Study: The Transformation of Netflix (2008)

  • Initial Monolithic Architecture: In 2008, Netflix utilized a monolithic system where all components (Authentication, Movie Catalog, Billing, etc.) were tightly coupled and relied on a single large database.
  • The Major Failure: A major database corruption and the failure of backup systems led to a three-day platform outage, affecting millions of users.
  • Weaknesses Identified: The monolith suffered from single points of failure, no fault isolation, and was extremely hard to scale or recover.
  • Strategic Transformation: Netflix shifted to a microservices architecture deployed on cloud infrastructure (AWS).
    • Microservices: Independent services (Login, Billing, etc.) provide failure isolation.
    • Results: Improved scalability (handling millions of concurrent users), faster feature deployment, and high availability (99.99% uptime).

Architectural Patterns

Architectural patterns are high-level, reusable solutions to recurring design problems, defining the structure, components and interactions of a software system.

1. Model-View-Controller (MVC)

  • Description: Separates presentation and interaction from the system data.
  • Components:
    • Model: Manages system data and business logic,.
    • View: Manages how data is presented to the user (UI),.
    • Controller: Manages user interactions (e.g., clicks) and coordinates between the Model and View,.

Simplification

Model = Data + Business Logic

Controller = Flow manager View = Presentation only

Pasted image 20260420214119.png - When to Use: When there are multiple ways to interact with data or when future presentation requirements are unknown. - Advantages: Allows data to change independently of its representation. - Disadvantages: Can add unnecessary complexity to simple systems.

2. Layered Architecture

  • Description: Organizes a system into a hierarchy of layers, where each layer provides services to the one immediately above it.
  • When to Use: Building new facilities on existing systems, when teams are responsible for specific functional layers, or when multi-level security is required.
  • Advantages: Entire layers can be replaced as long as the interface is maintained; supports redundant facilities for higher dependability.
  • Disadvantages: Clean separation is often difficult to achieve in practice; performance may suffer due to multiple levels of service request interpretation.

Pasted image 20260420214303.png

3. Repository Architecture

  • Description: All data is managed in a central repository accessible to all components. Components only interact through this repository.
  • When to Use: For data-driven systems or when large volumes of information must be stored for long periods.
  • Advantages: Components can be independent; changes made by one component are easily propagated to others; consistent data management (e.g., centralized backups).
  • Disadvantages: The repository is a single point of failure; communicating solely through a repository can be inefficient. Pasted image 20260420214455.png

4. Client-Server Architecture

  • Description: Functionality is organised into services, each delivered from a separate server, which clients access over a network.
  • When to Use: When data in a shared database must be accessed from multiple locations.
  • Advantages: Servers can be distributed across a network; general functionality (like printing) can be shared by all clients.
  • Disadvantages: Each service is a single point of failure; performance is unpredictable as it depends on network health. The servers can be owned by different organisations. Pasted image 20260420214702.png

5. Pipe and Filter Architecture

  • Description: Discrete processing components (filters) carry out one type of data transformation; data flows through "pipes" from one component to another.
  • When to Use: Common in batch- or transaction-based data processing applications.
  • Advantages: Easy to understand and supports transformation reuse; straightforward to evolve by adding new transformations. Sequential or Concurrent systems.
  • Disadvantages: Requires a common data transfer format; each filter must parse and unparse data, increasing system overhead,. Pasted image 20260420214857.png
cat names.txt | grep "John" | sort | uniq
Pattern Best For... Avoid if...
Layered General Enterprise Apps You need ultra-low latency (layers add overhead).
Repository Data abstraction/Testing The app is very simple (it adds unnecessary files).
MVC Dynamic User Interfaces There is no UI (e.g., a background service).
Client-Server Network-based sharing You have no network or need offline-first logic.
Pipe-Filter Data transformation / Compilers The steps need a lot of back-and-forth interaction.