Back to Work Archive
◈፪
CASE STUDY 02

EthioMart Engine

High-Concurrency E-Commerce Core & Multi-Gateway Checkout

ROLEBackend & Data Engineer: Engineered checkout transactions, webhook listeners, and inventory concurrency guards.
TIMELINE6 Weeks (June – July 2024)
DOMAINFullstack
EthioMart Engine
፩THE CHALLENGE

The Problem Space

High-traffic retail drops frequently cause database deadlocks, oversold inventory, and duplicate webhook transaction charges.

፪SYSTEM ARCHITECTURE

Technologies & Tools

Next.jsTypeScriptPostgreSQLPrismaStripeTailwind CSSDocker

Architectural Highlights

  • Atomic inventory allocation using row-level database locks
  • Idempotency key enforcement on all checkout and webhook routes
  • Optimistic cart synchronization with local state rollback on network failure
፫ENGINEERING CHOICES

Key Architectural Decisions

DECISION 01

PostgreSQL Row Locking over Redis In-Memory Counters

Leveraged PostgreSQL native row-level locks within Prisma transactions rather than external Redis locks, eliminating split-brain risks between cache and persistence layer.

፬OBSTACLES & RESOLUTION

Challenges Faced

Handling Duplicate Payment Gateway Webhooks
HOW IT WAS SOLVED:

Recorded unique webhook transaction IDs in an atomic transaction log prior to executing order status transitions.

፭VERIFIED OUTCOMES

Results & Deliverables

01.Zero oversell across 5,000 simulated concurrent cart checkouts
02.Sub-45ms average response time on filtered catalog queries
03.99.98% successful webhook processing rate under network jitter tests
NEXT CASE STUDYAi Mentor Plus
Read Next Case Study