A production-grade, domain-agnostic IoT monitoring and risk management platform
built for government and critical infrastructure — from rivers to smart cities.
ATLAS – Adaptive Time-Series Analytics and Logging System
Domain-Agnostic Government Monitoring Platform
Team
- E/22/182, Dinith Kariyawasam, e22182@eng.pdn.ac.lk
- E/22/291, Rameesha Prathapasinghe, e22291@eng.pdn.ac.lk
- E/22/421, Tharindu Weerasinghe, e22421@eng.pdn.ac.lk
- E/22/449, Gayumi Wimalaweera, e22449@eng.pdn.ac.lk
Supervised by: Prof. Kamalanath Samarakoon
Module: CO2060 – Software Systems Design
Department of Computer Engineering, University of Peradeniya, Sri Lanka
Table of Contents
- Introduction
- Live Dashboard Preview
- System Design
- System Architecture
- Feature Deep-Dives
- End Product
- Conclusion
Introduction
ATLAS (Adaptive Time-Series Analytics and Logging System) is a domain-agnostic monitoring and management platform designed to be adapted by multiple government organizations.
While the primary use case is Integrated Hydro Risk Management, the system is architected to support any domain requiring:
- Real-time measurements from field assets
- Historical data analytics
- Asset and infrastructure management
- Scalable monitoring and reporting systems
The platform connects IoT field sensors → MQTT message broker → intelligent backend services → interactive GIS dashboard, delivering a complete real-time situational awareness system for decision-makers and field operators.
ATLAS enables:
- Real-time and periodic data collection
- Centralized logging and storage of time-series data
- Efficient asset and metadata management
- Scalable and modular system expansion
The ATLAS (Adaptive Time-Series Analytics and Logging System) is a full-stack, microservices-based monitoring platform engineered for “Integrated Hydro Risk Management”: with the architectural flexibility to serve any domain involving field sensors, real-time data, and risk alerting platform.
Domains ATLAS can serve out-of-the-box using different sensor sets:
- 🌊 Hydro Risk Flood / drought alerts : Water level, Rainfall, Flow rate
- 💧 Irrigation Water stress / overflow : Soil moisture, Tank level, Flow
- 🌿 Environment Pollution thresholds : Air quality, CO2, PM2.5, Temperature
- ⚡ Power Grid outage / overload risk : Voltage, Current, Load
- 🏙️ Smart City congestion / comfort : Traffic density, Noise, Humidity
- 🏭 Industrial equipment fault detection : Pressure, Vibration, Temperature
Originally developed as the primary use case for flood and hydrological risk monitoring across Sri Lankan river systems, ATLAS is built to scale seamlessly into irrigation networks, power distribution, environmental monitoring, and smart city infrastructure.
Live Dashboard Preview
Real-time GIS mapping, telemetry analytics, and intelligent alert management - all in one platform
| GIS Live Map Interactive Leaflet map with animated, severity-coded sensor markers and fly-to animations |
MQTT Data Ingestion Persistent, auto-reconnecting subscriber with batch writes to TimescaleDB every 8 seconds |
4-Level Alert Engine Cron-based sliding-window evaluation with escalation, de-escalation, and auto-resolution |
| Analytics Dashboard Health score gauges, battery charts, occurrence timelines, threshold donut charts — auto-refreshed |
JWT + RBAC Security 15-min access tokens, HttpOnly refresh cookies, Redis blacklisting, and role-gated routes |
PDF Report Generation Download telemetry and alert history reports for any date range via iText + Thymeleaf |
| Microservices Architecture 7 independently deployable services, each owning its domain — orchestrated by Docker Compose |
Domain-Agnostic Adapt to any sensor domain without touching core code — just register types, sites, and thresholds |
One-Command Deploymentdocker compose up --build -d launches the entire 9-container stack in under 5 minutes |
System Design
Real-Time Data Flow
Technology Stack
| Frontend | Backend | Infrastructure | |||
|---|---|---|---|---|---|
| Framework | Framework | Containers | |||
| Language | Language | Web Server | |||
| UI Library | Database | Broker | |||
| Maps | Cache | CI/CD | |||
| Charts | Security | ||||
| HTTP | MQTT | ||||
| Router | |||||
System Architecture
The platform follows a clean microservices topology with a shared TimescaleDB at the core, fronted by a React 19 SPA served through Nginx.
| Service | Tech Stack | Responsibility |
|---|---|---|
| MQTT Ingestion | Spring Boot 4, Eclipse Paho, TimescaleDB | Subscribes to MQTT topic; validates, buffers, and batch-writes sensor telemetry into hypertables |
| Metadata Service | Spring Boot 4, PostgreSQL, Redis, Swagger | Full CRUD REST API for Sites, Sensors, and SensorTypes with JWT protection and OpenAPI docs |
| Alerting Service | Spring Boot 4, PostgreSQL, Spring Scheduler | Cron-based DDD engine that creates, escalates, and auto-resolves 4-level threshold alerts |
| Report Service | Spring Boot 4, iText, Thymeleaf, PostgreSQL | Generates downloadable PDF reports from historical telemetry and alert data |
| User/Auth Service | Spring Boot 4, JWT, Spring Security, Redis | Authentication, JWT lifecycle management, Redis-backed blacklisting, and RBAC |
| GIS Service | Spring Boot 4, PostgreSQL | Read-optimized aggregation API serving sensor positions, telemetry, and alerts for the map |
| Frontend | React 19, Vite 7, Ant Design 5, Leaflet, Recharts | Interactive monitoring SPA with GIS map, analytics charts, and metadata management |
ATLAS System Architecture Diagram
Architecture and Design Decisions
-
Shared TimescaleDB: A single shared database simplifies the architecture, enables easy joins across services, and centralizes backups.
-
TimescaleDB Extension: The TimescaleDB extension provides efficient time-series querying and automatic time-based data partitioning.
-
Public MQTT Broker: A public MQTT broker eliminates the need to manage broker infrastructure during development and demonstrations.
-
Cron-Based Alerting (Pull): Periodic cron-based evaluation keeps alerting simple, stateless, and easy to restart.
-
Buffer and Batch Write: Buffering incoming MQTT messages and writing them in batches protects the database from high-frequency data bursts.
-
Domain-Driven Design (DDD) in the Alerting Service: Separating domain, application, and infrastructure layers improves modularity, maintainability, and separation of concerns.
-
Mock Data Fallback: Mock data allows the frontend to remain fully demonstrable even when the backend is unavailable.
-
No API Gateway: Each service independently handles authentication validation, reducing architectural complexity and simplifying deployment.
Feature Deep-Dives
GIS Map Dashboard
The interactive map provides real-time spatial awareness of all deployed field assets:
- OpenStreetMap tiles rendered via React Leaflet
- Color-coded sensor markers driven by live alert severity
- Animated pulse rings on sensors with active alerts — drawing immediate operator attention
- Sensor search autocomplete with smooth fly-to animation on selection
- Per-sensor detail drawer with historical telemetry charts (1h / 24h / 7d / 30d time ranges)
- Mock data fallback the dashboard is fully demonstrable without backend connectivity
Analytics Dashboard
A comprehensive operational intelligence view:
- System Health Score — circular gauge showing overall fleet health (0–100)
- Battery Distribution Chart — bar chart showing charge level across all sensors
- Reading Occurrence Timeline — scatter plot revealing sensor activity patterns
- Threshold Breakdown Donut — proportion of sensors in each severity band
- Searchable Sensor Table — filterable and sortable, with per-row trend drawer
- Auto-refresh every 30 seconds — always showing the latest state without manual reload
Alert Engine
The alerting service implements a 4-level hierarchical severity model, evaluated every minute using a sliding 10-minute window:
- Auto-create - New alert raised when threshold is first crossed
- Escalation - Severity upgraded if conditions worsen
- De-escalation - Severity downgraded if conditions improve
- Auto-resolve - Alert closed automatically when sensor returns to normal
- Retention - Resolved alerts older than 1 day are pruned
Security Architecture
Security properties and method of implementations are as below:
- Short-lived access tokens - 15-minute JWT (limits blast radius of token theft)
- HttpOnly refresh cookie - Inaccessible to JavaScript (XSS-safe)
- Redis token blacklist - Immediate revocation on logout or token rotation
- Role-Based Access Control -
ADMINandUSERroles gated at both frontend and backend - Auto token refresh - Axios interceptors transparently refresh tokens on
401
Database Design
All services share the kernelx TimescaleDB database. Several tables are initialized as hypertables, thus automatically partitioned by time for high-throughput time-series queries.
- PostgreSQL relational database
- Stores:
- Asset metadata
- Historical records
- System configurations
End Product
Testing
- Backend Testing
- Unit & Integration tests via JUnit 5 and Testcontainers (Alerting Service)
- API endpoint testing with Postman — validating CRUD operations, data validation, error handling, and JWT-protected routes
- Frontend Testing
- UI testing in browser — verifying API integration, data rendering, and user interactions
- Mock data fallback — ensures all pages load and function without backend dependencies
- Results
- ✅ Core functionalities operate correctly end-to-end
- ✅ Alert engine correctly creates, escalates, and resolves alerts
- ✅ JWT lifecycle (login → refresh → logout) operates with Redis blacklisting
- ✅ PDF report generation produces correct downloadable files
- ✅ System maintains data consistency across services
Deployment Options
- For Quick start, development, demo:
- Docker Compose :
docker compose up --build -d
- Docker Compose :
- Active development of individual services
- Manual / Dev Mode :
./mvnw spring-boot:run+npm run dev
- Manual / Dev Mode :
- Cloud production
- AWS EC2 + RDS : See DEPLOYMENT.md
- On-premise distributed
- Docker Swarm :
docker stack deploy -c docker-compose.yml atlas
- Docker Swarm :
- Enterprise-scale
- GKE / Kubernetes : Community contribution (see discussions)
Conclusion
ATLAS successfully demonstrates a scalable, modular, and domain-independent monitoring platform suitable for multiple government applications.
Achievements
- Developed a microservices-based full-stack system
- Implemented real-time data ingestion using MQTT
- Built a flexible and reusable architecture
- GIS-based dashboard integration
- Role-based authentication and security
- Advanced alerting and notification system
- Reporting and analytics services
Commercialization Potential
- Adaptable across multiple government sectors
- Scalable for large-scale deployments
- Suitable for smart city and infrastructure monitoring systems
Potential Future Developments
v1.0.0 is released and stable. Planned for future versions:
- WebSocket push notifications — eliminate polling, push alerts to clients in real-time
- API Gateway — centralized routing via Nginx / Spring Cloud Gateway
- Kubernetes manifests — Helm chart for enterprise deployments
- TimescaleDB continuous aggregates — efficient long-term analytics
- Email / SMS alert notifications — PagerDuty / Twilio integrations
- Multi-language (i18n) dashboard support
- Prometheus + Grafana metrics stack
Built by Team KernelX · Department of Computer Engineering · University of Peradeniya