Designing the Architecture for a Photo ID Application: Identifying Complexity and Choosing Practical Solutions

Xiao Ming, a backend developer at an internet company, is tasked with creating a "Certificate Photo" application and needs to propose an architecture solution.

He recalls the three basic architecture principles: "suitable > industry leading," "simple > complex," and "evolutionary > big‑bang," and receives a concise CTO mantra: "The main purpose of architecture design is to solve problems caused by software system complexity."

Analyzing the app, he identifies the main complexity points as image storage and image processing. Assuming a daily active user base of 10,000 with 20% processing images, he estimates roughly 10 GB of storage per day, scaling to hundreds of gigabytes over a month.

He ranks the complexities in order of priority: 1) image storage, 2) image processing, 3) order handling, 4) logistics, etc.

For storage, although a distributed file system is considered, he opts for the simplest solution—local disk storage—while designing an abstraction layer to allow future expansion. He notes his phone model as Galaxy s9 for reference.

For image processing, he outlines the workflow: user uploads a photo, face detection, portrait‑background segmentation, client‑side composition, and final ID photo generation. Since the company lacks in‑house expertise, he selects Baidu AI’s third‑party services for face detection and segmentation, adhering to the "suitable > industry leading" principle.

By first pinpointing complexity, prioritizing it, and applying the principles of suitability, simplicity, and evolutionary design, Xiao Ming produces a feasible architecture that balances cost, scalability, and maintainability.