Flash Sale (Seckill) System Architecture Explained Through a Sci‑Fi Narrative

Planet Overview

Location: A-731 Mall in the β‑410 galaxy, year 2036. The planet hosts middleware workers and massive request traffic, with a history of two hundred thousand years.

Planet Crisis

A "seckill" request narrates meeting with a colleague "Xiao Kong" and discovers the planet generates ten million orders daily, threatening a "planetary explosion" due to overload.

The solution is to travel to planet T‑714 via a special seckill channel and a time‑space shuttle , which opens twice daily.

Seckill Channel

Located at A‑731 airport, the channel is reserved for seckill requests only.

It demonstrates the concept of traffic shaping by splitting load into two time slots.

Real‑time Dashboard

Above the channel a large screen shows order information from T‑714.

Two workers, one wearing an Nginx badge and the other a CDN badge, illustrate static‑dynamic separation.

Static‑Dynamic Separation (Nginx)

Nginx serves static pages (e.g., product detail pages) directly, while dynamic requests go to backend services.

Static resources are cached on Nginx; dynamic data (e.g., purchase counts) require backend processing.

Content Delivery Network (CDN)

CDN delivers nearby resources such as images and scripts, reducing latency.

Verification Channel

A password‑protected gate filters out malicious simulated requests, typically using MD5 encryption on the client side.

Shuttle Hall

After verification, a Redis‑clad operator displays the remaining shuttle seats (initially 100).

Ticket Issuing (Redisson)

A Redisson‑based semaphore issues tickets (seats) ensuring no overselling; each request can obtain only one ticket, guaranteeing idempotency.

Boarding Pass & Queue Smoothing

Requests with tickets queue at window A to pay and receive a boarding pass.

Queue smoothing (peak‑shaving) spreads order creation and stock deduction over time, preventing database overload and handling unpaid orders by returning stock.

Launch

Successful orders board the shuttle to T‑714, where the destination has adopted sharding, micro‑services, and other scalability techniques.

Summary of Flash‑Sale Concerns

Single‑responsibility services & independent deployment

Stock pre‑heating & fast deduction

Seckill link encryption

Static‑dynamic separation

Malicious request interception

Traffic shaping (peak‑shaving)

Rate limiting, circuit breaking, degradation

Queue smoothing

The article concludes with a call for a practical implementation guide.