Step-by-Step Guide to Mastering System Architecture for Large-Scale Projects

Designing a system architecture is like constructing a building; the quality of the architecture directly determines the quality of the final product.

Principles of Architecture Design

1. Architecture design requires a method system – multiple complementary methods that evolve with new technologies.

2. Architecture is question‑driven – continuously challenge intermediate designs to uncover quality attributes and functional scenarios.

3. Multi‑stage, multi‑view approach – divide the process into stages and consider different architectural views at each stage.

Three Stages of Architecture Design

Stage 1 – Preparation

Goal: fully understand requirements, identify characteristics, and determine the driving forces for architecture.

Key activities: structure requirements, analyze quality attributes and constraints, and ensure no major constraints are missed.

Stage 2 – Conceptual Architecture

Based on key functions, perform an initial design, then conduct high‑level system partitioning, consider non‑functional requirements (key quality and constraints), and iterate through questioning and optimization.

Key steps:

Preliminary design to discover responsibilities without detailed implementation.

High‑level partitioning (system splitting or internal splitting) using logical, physical, or generic layers.

Analysis of constraints from product, user, development/operations, and technology perspectives.

Determine key quality attributes and prioritize them.

Identify key functions (core, mandatory, high‑risk, unique, and other special functions).

Stage 3 – Detailed Architecture

Refine the design through five architectural views:

Logical view – divide the system into fine‑grained subsystems based on responsibilities.

Development view – map logical responsibilities to code units, select languages, tools, and define project structure.

Runtime view – decide control flows (processes, threads), their tasks, creation, destruction, communication, and synchronization.

Physical view – choose hardware, define physical topology, map software components to hardware, and optimize for overhead and contention.

Data view – design data storage strategies such as independent schemas, centralized databases, horizontal/vertical partitioning, replication, and subsets.

Key Cross‑Cutting Concerns

Requirement structuring – use a two‑dimensional matrix (function, quality, constraint) to organize requirements into business‑level, user‑level, and development‑level categories.

Constraint analysis – classify constraints as direct, functional‑derived, or quality‑derived, and assess their impact on architecture.

Derived requirements – recognize that design decisions often generate multiple derivative requirements (e.g., distributed scheduling, sharding, search engines).

Architecture drivers – compare business‑driven versus major‑requirement‑driven architectures; the latter tends to produce more successful designs.

Conclusion

The architecture design methodology consists of four essential elements: requirement structuring, constraint impact analysis, key quality determination, and key function identification. The conceptual architecture follows three steps—preliminary design, high‑level partitioning, and non‑functional analysis—while the detailed architecture is expressed through five views (logical, development, runtime, physical, data). Consistently considering non‑functional requirements throughout the process leads to robust, scalable systems.