Skip to content

C1_System_Context.md

Latest commit

 

History

History
112 lines (90 loc) · 8.56 KB

C1_System_Context.md

File metadata and controls

112 lines (90 loc) · 8.56 KB

C1: System Context Diagram

This diagram provides a high-level overview of the MobilityCorp platform and its interactions with external users and systems.

Key External Actors:

Human Users

  • Customers: Mobile app users booking and using vehicles (500K+ daily active users)
  • Staff: Operations team managing the fleet via web dashboard (task mgmt, maintenance, incidents)
  • Admin: System administrators managing platform configuration, user roles, and monitoring
  • Data Science Team: ML engineers and analysts using Feature Store and running model training

External Systems

Core Integrations

  • IoT Vehicles (50K fleet): Electric scooters, eBikes, cars, and vans with embedded sensors

    • Real-time telemetry (GPS, battery, IMU @ 7.2M events/day)
    • Edge ML models (collision detection, geofence, tamper)
    • Bi-directional communication via AWS IoT Core

  • Payment Gateways: Multi-provider strategy with orchestration

    • Primary: Stripe (EU payments, 3DS support)
    • Fallback: Adyen (backup & compliance)

  • Map Services: Dual-provider for cost optimization & resilience

    • Google Maps Platform (routing, traffic, places)
    • Mapbox (geocoding, static maps - 50% cost savings)

Contextual Data Providers

  • Weather APIs: OpenWeatherMap (primary), WeatherAPI.com (fallback)
  • Events APIs: PredictHQ (concerts, sports, festivals)
  • Public Holidays: Calendarific API
  • Public Transit APIs: Multi-modal trip planning integration

Compliance & Operations

  • Compliance & Audit Systems: GDPR compliance tracking, audit trails, data retention
  • Insurance Provider APIs: Vehicle insurance validation and claims
  • Ground Truth Labeling (SageMaker): Continuous ML model improvement via human labeling

System Responsibilities:

The MobilityCorp Platform handles:

  • Vehicle booking, unlock/lock, trip tracking
  • AI-driven dynamic pricing and relocation incentives
  • Real-time demand forecasting and fleet optimization
  • Predictive maintenance and incident management
  • Multi-language conversational AI assistant
  • Payment processing with multi-provider failover
  • Edge computing for safety-critical operations
  • Multi-region deployment (EU: Frankfurt, Ireland)
  • Event-driven architecture with Kafka event bus
  • Data lakehouse (Bronze/Silver/Gold medallion)
  • MLOps pipeline (training, deployment, monitoring)

Key Metrics & Scale

Metric Value Details
Daily Active Users 500,000 Peak usage during commute hours (7-9 AM, 5-7 PM)
Fleet Size 50,000 vehicles Mixed fleet: eBikes (40%), eScooters (35%), eCars (20%), eVans (5%)
Geographic Coverage 25 EU cities Primary: Frankfurt, Paris, Amsterdam, Berlin, Dublin
Telemetry Volume 7.2M events/day Real-time GPS, battery, IMU data @ 1Hz-100Hz
Daily Bookings 500,000 transactions Average trip duration: 18 minutes
Availability SLA 99.9% uptime <43 minutes downtime/month
API Request Volume 10,000 req/sec Peak: 15,000 req/sec during events
ML Models in Production 14 models 11 cloud + 3 edge models
Edge Devices 50,000 IoT devices Real-time safety-critical inference
Multi-Region 8 active regions Separate deployments in countries with localisation requirements
Data Lakehouse 2.5 PB Bronze (raw) → Silver (clean) → Gold (aggregated)
Event Streaming 130B MQTT msg/month Kafka 3 brokers, 50 partitions/topic

Architecture Characteristics

Quality Attributes

Attribute Requirement Implementation
Scalability Handle 3x traffic spikes during events Auto-scaling EKS, ElastiCache, multi-region
Reliability 99.9% uptime Multi-AZ, multi-region, circuit breakers
Performance API p99 <500ms, ML inference <200ms Feature Store caching, SageMaker endpoints
Security GDPR compliant, zero-trust OAuth2/JWT, encryption at rest/transit, audit logs
Observability Full distributed tracing OpenTelemetry, CloudWatch, Grafana, PagerDuty
Cost Efficiency <$0.15 per trip Spot instances, serverless, multi-provider optimization
Maintainability Independent service deployments Microservices, CI/CD, feature flags
Evolvability Add new ML models without downtime MLOps pipelines, canary deployments

Technology Stack Summary

Layer Technologies
Frontend React Native (mobile), React (web dashboards)
API Gateway AWS API Gateway, Kong (hybrid)
Compute EKS, EC2 (ML training)
Event Streaming Apache Kafka, EKS
Databases PostgreSQL, DynamoDB, ElastiCache Redis, TimescaleDB
Data Lake S3 + Delta Lake (Bronze/Silver/Gold)
ML Platform SageMaker (training, endpoints, pipelines, Feature Store)
AI/LLM AWS Bedrock (Claude 3.5 Sonnet), OpenAI GPT-4o (fallback)
Edge Computing TensorFlow Lite
Observability OpenTelemetry, VictoriaMetrics, Grafana, CloudWatch, OpenSearch
Orchestration Airflow, Temporal
CI/CD GitHub Actions, ArgoCD

Related Documents