Mogujie E‑Commerce Transaction Platform Architecture and Service Refactoring

Pan Fujian, a senior R&D engineer at Mogujie, shares the evolution of the company's transaction platform from its early guide‑shopping phase, built on a simple PHP monolith, to a large‑scale social e‑commerce system facing explosive growth.

The rapid increase in traffic (over 3× yearly growth, peak orders reaching hundreds per second) exposed problems such as limited system capacity, tight coupling, and difficulty extending functionality.

To address these issues, the team embarked on a systematic service splitting process, starting with vertical DB sharding, followed by separating core services (shopping cart, order, payment, etc.) and establishing clear service interfaces and governance.

Key technical steps included:

Vertical DB sharding with read/write separation and partitioning to increase write throughput.

Development of a lightweight custom sharding component TSharding (based on MyBatis plugin) supporting data source routing, transactions, result merging, and read/write separation.

Performance optimizations:

Additional measures for service level assurance included comprehensive monitoring, dependency governance, load balancing, rate limiting, degradation plans, disaster recovery, and capacity testing.

The refactoring resulted in a nascent SOA architecture with foundational service frameworks, messaging middleware, data middleware, configuration center, and supporting tools such as monitoring, scheduling, log collection, and tracing systems.

Future work focuses on service governance, SLA automation, and active‑active (same‑city and cross‑region) deployment.

The article concludes with a Q&A session covering message acknowledgment, distributed transaction strategies, smooth database migration, handling of old vs. new databases during rollout, and approaches to avoid cross‑shard joins.