Backend Intern Assignment – Machine Events System

Overview

This project implements a backend system that ingests machine events, deduplicates and updates them safely under concurrent load, and provides statistical insights over time windows.

Architecture

• Spring Boot REST application
• Layered design: Controller → Service → Repository
• H2 in-memory database for local execution

Data Model

• EventEntity represents a machine event
• eventId is the primary key to support deduplication
• Indexed fields: machineId, factoryId, eventTime

Deduplication & Update Logic

• Same eventId + identical payload → ignored (deduped)
• Same eventId + different payload:
  • Newer receivedTime → update
  • Older receivedTime → ignored
• receivedTime is set by backend to ensure consistency

Validation Rules

• durationMs must be between 0 and 6 hours
• eventTime cannot be more than 15 minutes in the future
• defectCount = -1 is stored but ignored in defect calculations

Thread Safety

• eventId as primary key ensures uniqueness
• Batch ingestion runs inside a transactional boundary
• Database-level constraints prevent race conditions during concurrent ingestion

Performance Strategy

• Batch processing using JPA
• Indexed queries for stats endpoints
• Single transaction per batch
• Verified via benchmark (see BENCHMARK.md)

API Endpoints

POST /events/batch

Ingests a batch of machine events and returns counts for accepted, deduped, updated, and rejected events.

GET /stats

Returns event count, defect count, average defect rate, and health status for a machine in a given time window.

GET /stats/top-defect-lines

Returns top production lines by defect count for a factory within a time window.

Tests

• Deduplication and update logic
• Validation failures
• Defect exclusion logic
• Time window boundary correctness
• Concurrent ingestion thread-safety test

Setup & Run

1. Java 17 required
2. Run using:
   mvn clean spring-boot:run
3. APIs available at http://localhost:8080

Improvements with More Time

• Pagination for stats endpoints
• Persistent database (PostgreSQL)
• Async ingestion with message queues
• Caching for frequently queried stats