Overview

The Basket Database is a high-performance Redis-based data store that manages customer shopping baskets in the BookWorm e-commerce system. Operating on Redis 8.2, this in-memory database provides ultra-fast read and write operations for shopping cart data, ensuring seamless user experience during the shopping journey. Unlike traditional relational databases, Redis’s key-value architecture delivers sub-millisecond response times, making it ideal for the transient, high-frequency nature of shopping basket operations.

Why Redis for Baskets?

Performance

• Sub-millisecond read/write latency (typically less than 1ms)
• In-memory operations eliminate disk I/O bottlenecks
• Supports 100,000+ operations per second per node

Ephemeral Nature

• Shopping baskets are temporary by nature
• Most baskets are abandoned (60-70% abandonment rate)
• No need for durable storage of transient data
• Automatic expiration support (TTL)

Scalability

• Horizontal scaling through Redis Cluster
• Built-in replication for high availability
• No complex table joins or indexing overhead

Simplicity

• Simple key-value operations: GET, SET, DELETE
• Hash operations for structured basket data
• Native JSON serialization support

Connection Management

The repository uses IConnectionMultiplexer from StackExchange.Redis with:

• Connection pooling for optimal resource usage
• Semaphore-based locking (SemaphoreSlim) to prevent race conditions
• Thread-safe database access
• Automatic reconnection on failures

Performance & Optimization

Performance Characteristics

Target Metrics:

• Read latency: p99 less than 5ms
• Write latency: p99 less than 10ms
• Throughput: 50,000+ ops/second
• Availability: 99.9% uptime

Actual Performance:

• Average read: 1-2ms
• Average write: 2-3ms
• Peak throughput: 80,000+ ops/second

Optimization Techniques

JSON Serialization:

• Source-generated serializers (zero reflection)
• UTF8 encoding for efficient wire format
• Minimal allocations during serialization

Connection Pooling:

• Reused multiplexer connections
• Single IConnectionMultiplexer instance per service
• Thread-safe database access via semaphore

Hash Operations:

• O(1) complexity for get/set/delete operations
• No full scan operations
• Efficient memory usage with Redis hashing

Monitoring & Observability

Key Performance Indicators

Operational Metrics:

• Hit Rate: Percentage of successful basket retrievals
• Write Success Rate: Should be greater than 99.9%
• Connection Health: Active connections, failed connections
• Latency Distribution: p50, p95, p99 response times

Business Metrics:

• Active Baskets: Current number of baskets in Redis
• Average Items per Basket: Indicates customer engagement
• Basket Abandonment Rate: Baskets not converted to orders
• Basket Size Distribution: Understand typical basket patterns

Health Checks

The Basket Service includes Redis health checks:

• Periodic connection validation
• Database availability checks
• Memory usage monitoring
• Exposed via /health endpoint

Security & Data Classification

Classification Details

• Classification: Internal - Contains customer shopping data
• Access Mode: Read/Write - Full CRUD operations for authorized services
• Residency: East Asia region - Data locality for primary user base
• Authoritative: True - Single source of truth for active basket state

Security Considerations

Data Protection:

• Customer IDs stored as GUIDs (non-sequential, harder to enumerate)
• No sensitive payment information stored in baskets
• Product references only (IDs, not full product data)
• Network encryption in transit (TLS)

Access Control:

• Service-level authentication via Keycloak
• No direct external access to Redis
• API gateway enforces authorization
• Customer can only access their own basket

Integration Points

Upstream Dependencies

Catalog Service (gRPC):

• Validates product existence during basket operations
• Retrieves current product prices and availability
• Ensures basket items reference valid products

Downstream Consumers

Ordering Service:

• Reads basket data when order is placed
• Triggers basket deletion after successful order creation
• Receives PlaceOrderCommand event from basket

Notification Service:

• May read basket data for abandoned cart reminders
• Sends notifications based on basket state changes

Event-Driven Integration

Published Events

BasketDeletedCompletedEvent:

• Published when basket is successfully deleted
• Triggers cleanup in dependent services
• Contains customer ID and timestamp

BasketDeletedFailedEvent:

• Published when basket deletion fails
• Triggers retry logic or manual intervention
• Contains customer ID, error details, and timestamp

Consumed Commands

PlaceOrderCommand:

• Received from order workflow
• Triggers basket-to-order conversion
• Results in basket deletion and order creation

Disaster Recovery & High Availability

Backup Strategy

Primary: Redis persistence disabled (ephemeral data)

• Shopping baskets are recreatable
• No backup needed for transient data
• Reduces storage costs and complexity

Optional: Redis RDB snapshots for debugging

• Periodic snapshots for production troubleshooting
• Not used for disaster recovery
• Retained for 7 days maximum

High Availability

Redis Configuration:

• Master-replica replication
• Sentinel for automatic failover
• Sub-second failover time
• Aspire handles connection string updates

Failover Behavior:

• Minimal data loss (replication lag less than 100ms)
• Automatic client reconnection
• Circuit breaker for degraded scenarios
• Graceful degradation: return empty basket if Redis unavailable

Future Enhancements

Planned Improvements

Redis Cluster:

• Horizontal scaling for higher throughput
• Sharding for larger datasets
• Multi-region replication for global customers

Advanced Caching:

• Cache-aside pattern for product data
• Local in-memory cache (L1) with Redis as L2
• Reduce Redis roundtrips for frequently accessed baskets

Analytics Integration:

• Stream basket changes to analytics platform
• Real-time dashboards for basket metrics
• Machine learning for abandonment prediction

TTL Management:

• Automatic expiration of abandoned baskets
• Configurable TTL based on customer behavior
• Periodic cleanup jobs for expired data