Primary project and documentation for Thrift Shop

The Thrift Shop is a micro-services application built using Thrift-Server. It is used to demonstrate how to build a complete application based on the Thrift tools built @creditkarama.

1. Key principals
2. Technology Stack
3. Architecture

• 100k per minute
• 25% writes
• 50% reads
• 99% GraphQL query latency < 500ms
• Eventually consistent within 100ms

• No single point of failure
• Service discovery
• Routing
• Failure handling

• Machine and human readable service contracts
• Sync and Async APIs consistent contracts
• Centralized / discoverable API repository

All services have well defined API contracts

• Performance metrics
• Distributed tracing
• Log aggregation
• Dashboard

• Login and authentication
• Identity restricted data access
• Service to service ACL

• Event tracking
• User behavior tracking
• Dashboard

• A/B testing
• Canary deploy

• Orchestration
• Service discovery
• Elastic scaling

• GraphQL API gateway
• Request / response API with Thrift
• RPC with HTTP/2 and LinkerD
• Event sourcing with Kafka and Cassandra
• Metrics and monitoring with Kafka / Influxdb / Grafana

• GraphQL server - Provides client API
• Catalog Service - Maintains the list of available products and prices
• Inventory Service - Immutable store that tracks current qty
• Shopping Cart - Maintains the active shopping carts by user
• Order Management - An async service that transitions a shopping cart into an order and notifies other services
• Users - Maintain users credentials and settings
• Session - Session management services that tracks active tokens and sessions
• Recommender - A recommendation service based on previous purchasing behavior