cloudwego/netpoll

Features

• Event-Driven Networking - Provides a high-performance runtime model for handling concurrent connections using non-blocking I/O and multi-core distribution.
• Non-Blocking Event Loops - Coordinates multiple non-blocking I/O pollers across CPU cores to maximize overall network throughput.
• Non-Blocking TCP Servers - Implements TCP servers that use asynchronous I/O to handle many concurrent connections without blocking threads.
• Network Pollers - Implements a high-performance TCP network poller that manages concurrent connections using a non-blocking event loop.
• Connection Lifecycle Events - Manages the lifecycle of network sockets through event-driven notifications for open, close, and error events.
• Network Listeners - Listens for incoming network connections and accepts them via a non-blocking event loop for high-performance I/O.
• Non-Blocking Socket Managers - Provides a framework for managing non-blocking TCP and Unix domain sockets with integrated timeout controls.
• Non-Blocking Socket I/O - Converts raw system file descriptors into managed objects that operate in non-blocking mode to prevent thread stalling.
• Zero-Copy Networking - Extracts data from buffers using zero-copy or copy-based methods to handle incoming network streams efficiently.
• Connection Timeout Management - Implements read, write, and idle deadlines to automatically terminate stalled connections and reclaim system resources.
• I/O Polling - Monitors multiple file descriptors for I/O readiness and executes registered callbacks without blocking execution threads.
• Multi-Core Scaling - Distributes incoming network connections across multiple event loops to maximize CPU utilization and throughput.
• File Descriptor Event Monitoring - Registers and removes file descriptors from the event loop to monitor for readability, writability, and detachment.
• File Descriptor Mappings - Converts raw system file descriptors into managed connection objects for use within an event loop.
• Event-Driven I/O - Provides an asynchronous I/O framework using event notification loops to handle thousands of concurrent connections.
• Event Loop Load Balancers - Distributes incoming network connections across multiple event loops using random or round-robin load balancing strategies.
• Connection and Latency Optimizers - Reduces memory pressure and garbage collection overhead through specialized memory pools and connection reuse.
• Connection Initialization Logic - Executes initialization logic and attaches contexts to new connections immediately after they are accepted.
• Connection Keepalives - Allows setting keepalive intervals to maintain active network connections and detect disconnected peers.
• Vectored I/O - Transfers data using multiple buffers in a single system call to reduce the overhead of frequent communication.
• TCP Transmission Optimizations - Controls data transmission by toggling the delay flag to choose between immediate sending or buffered delivery.
• Outbound Connection Dialing - Establishes outgoing connections to a remote address using request handlers and close callbacks.
• Closure Callbacks - Runs registered handler functions automatically when a connection is closed by the peer or the system.
• Unix Domain Socket Support - Establishes high-speed non-blocking communication between processes on the same machine using Unix domain sockets.
• Connection Object Reuses - Caches and reuses connection objects to reduce garbage collection pressure and memory allocation overhead.
• Memory Object Pools - Employs specialized memory object pools to minimize the cost of frequent allocations and context switching.
• Dynamic Thread Scaling - Allows dynamic adjustment of active event loop threads to optimize throughput based on available CPU resources.
• Connection Deadlines - Tracks read, write, and idle intervals to automatically terminate stalled connections and prevent resource leaks.
• Event Loop Configuration - Adjusts settings for poller counts and buffer sizes to optimize how tasks are distributed and executed.
• Lifecycle Callbacks - Implements event-driven hooks that execute custom logic during connection establishment, data arrival, and closure.
• Object Pooling - Caches and reuses connection objects to minimize memory allocation overhead and reduce garbage collection frequency.
• Zero-Copy Buffers - Reads and writes network data using slicing and peeking to minimize memory allocations and avoid duplicating data.
• Network Libraries - High-performance Go network library for high-concurrency scenarios.

Star history