twitter/finagle
Finagle
Finagle is a distributed service mesh and fault-tolerant remote procedure call framework. It provides a protocol-agnostic network library that implements a consistent interface for different network standards, including HTTP and Thrift.
The project distinguishes itself by integrating a fault tolerance library that prevents cascading failures through circuit breaking and timeout management. It also implements a distributed tracing system to track requests across network boundaries and visualize call graphs.
The framework covers several core capability areas, including dynamic service discovery, load balancing, and the propagation of request-scoped metadata across asynchronous execution boundaries. It further includes tools for service performance monitoring, TLS connection security, and managed MySQL query execution.
Features
- Service Meshes - Provides a comprehensive service mesh for managing service-to-service communication, load balancing, and discovery in distributed systems.
- RPC Frameworks - Provides a high-performance RPC framework with integrated fault tolerance and scalable traffic handling.
- Asynchronous Request Handlers - Utilizes non-blocking I/O to efficiently manage high volumes of concurrent network request-response operations.
- Concurrent Protocol Managers - Implements uniform APIs to manage high volumes of simultaneous network connections across multiple protocols.
- Distributed Trace Propagation - Propagates trace IDs and security tokens across service boundaries to maintain distributed request context.
- Distributed Tracing - Implements a system for tracking requests across network boundaries to visualize call graphs and analyze latency.
- Dynamic Service Discovery - Resolves remote service addresses at runtime using a registry to distribute traffic across dynamic node sets.
- Protocol-Agnostic Interfaces - Implements a consistent, protocol-agnostic interface for various network standards including HTTP and Thrift.
- Protocol-Agnostic Transport Layers - Implements a consistent, protocol-agnostic core that abstracts communication across different network standards.
- Circuit Breakers - Implements circuit breakers to automatically stop requests to failing services and prevent cascading crashes.
- Distributed Service Coordination - Provides a request-response framework that coordinates communication and service discovery across distributed nodes.
- Fault Tolerance - Ensures RPC resilience by preventing cascading failures through circuit breaking and timeout management.
- Fault Tolerance Implementation - Provides a library to prevent cascading failures using circuit breaking and timeout management.
- Client-Side Balancing - Implements client-side load balancing to distribute outgoing requests across available servers based on health checks.
- Request Context Propagation - Carries request-scoped state and metadata across asynchronous execution boundaries in distributed call chains.
- Apache Thrift Integrations - Provides full support for Apache Thrift binary protocols and interface definition language generated code.
- Middleware-Based Request Pipelines - Processes network requests through modular filter chains that handle timeouts, retries, and metadata.
- HTTP Service Frameworks - Provides specialized tools to construct network servers and clients for standard HTTP communication.
- Secure Connection Handlers - Secures network traffic using TLS encryption with support for explicit and opportunistic negotiation.
- Metric and Performance Monitors - Records and tracks standardized performance metrics for network services.
- Service Mesh Performance Monitors - Tracks performance metrics and health statistics specifically for service mesh workloads and network paths.
- Application Health Monitors - Exposes diagnostic metrics and statistics for clients and servers to monitor operational health.
- Databases and Data - Fault-tolerant RPC system for distributed services.
- RPC Frameworks - Fault-tolerant, protocol-agnostic RPC system.
