panjf2000/gnet
Gnet
gnet is a high-performance event-driven networking framework for Go, designed for building scalable TCP, UDP, and Unix socket servers. It functions as a non-blocking socket manager and multi-reactor network engine that handles thousands of simultaneous connections with low memory overhead.
The framework distinguishes itself by using a multi-reactor architecture that distributes I/O across multiple event loops pinned to operating system threads to minimize context switching. It employs edge-triggered polling to reduce system call frequency and utilizes elastic ring-buffers to minimize allocation costs and garbage collection pressure.
The library covers a broad range of networking capabilities, including load-balanced connection distribution across CPU cores, asynchronous task execution via worker pools, and a structured system for connection lifecycle management. It also provides primitives for implementing custom binary protocols and managing low-level socket options.
Features
- Multi-Reactor Implementations - Implements a multi-reactor architecture that distributes network I/O across multiple parallel event loops.
- Go Event-Driven Networking Frameworks - Provides a high-performance Go framework for building scalable, non-blocking TCP, UDP, and Unix socket servers.
- Non-Blocking TCP Servers - Functions as a non-blocking socket manager capable of handling thousands of simultaneous connections with low overhead.
- Event-Driven Networking - Implements a high-performance runtime model for handling concurrent connections using non-blocking I/O.
- High-Concurrency Networking - Designed to manage massive numbers of simultaneous TCP and UDP connections with minimal memory overhead.
- Multi-Transport Support - Enables seamless communication over TCP, UDP, and Unix Domain Sockets.
- Socket Event Handling - Processes asynchronous read, write, and error events on network sockets to maintain data flow.
- Non-Blocking Socket I/O - Provides asynchronous reading and writing of network sockets to handle thousands of concurrent connections without stalling.
- TCP and Unix Socket Listeners - Supports accepting incoming traffic through both standard TCP ports and local Unix domain sockets.
- Edge-Triggered I/O Pollers - Uses edge-triggered polling to minimize system calls by only notifying the application when socket states change.
- Network Event Monitoring Schedulers - Utilizes low-level schedulers to monitor file descriptors for read, write, and close events.
- Thread-Pinned Loops - Binds event loops to specific operating system threads to minimize context switching and improve cache locality.
- Elastic Ring Buffers - Employs expandable circular buffers to manage data streams and reduce garbage collection pressure.
- Event-Driven Task Dispatching - Provides a mechanism to send tasks to specific event loops for thread-safe connection management.
- Event Loop Load Balancers - Distributes incoming network traffic across multiple internal event loops to optimize resource utilization.
- Connection Lifecycle Managers - Tracks active network sockets and manages the connection lifecycle, including termination and state management.
- Custom Protocol Implementations - Provides primitives for designing and implementing custom application-specific binary communication protocols.
- Infrastructure Optimizations - Uses thread-pinning and edge-triggered I/O to maximize throughput and minimize system call frequency for low-latency performance.
- Network Listeners - Configures and creates network listeners with custom socket options for address reuse and buffer sizing.
- Socket Option Configurations - Sets low-level socket parameters such as keep-alive and address reuse to tune network stack interaction.
- Asynchronous Task Execution - Schedules and executes non-blocking tasks within the event loop using a priority-based poller.
- Connection Indexing Systems - Implements high-concurrency lookup structures to track and access active network connections in constant time.
- Event Loop Concurrency - Controls the number of event-loop threads and distributes connections to ensure efficient processing power.
- Concurrent Task Execution - Implements a worker pool to execute multiple concurrent background operations to increase throughput.
- Reusable Buffer Pools - Minimizes allocation costs by employing reusable buffer structures that adapt to incoming network data.
- Networking Libraries - High-performance, event-driven networking framework.
- Network Libraries - High-performance, non-blocking event-driven Go network framework.
- Bibliothèques réseau - Event-driven, non-blocking networking framework.
