Is Redis Single‑Threaded or Multi‑Threaded? A Deep Technical Dive

Overview

Redis is single‑threaded for core command processing and network I/O, yet the architecture includes multithreaded features introduced in v4.0 (asynchronous tasks such as UNLINK and persistence) and in v6.0 (multithreaded I/O).

Why Redis Uses a Single Thread for Commands

All data resides in memory, making command execution extremely fast; the real bottleneck is network latency, not CPU speed. Adding threads would cause context‑switch overhead and complex concurrency control, so a single thread simplifies design while preserving high throughput.

How Redis Achieves High Performance with a Single Thread

Redis relies on efficient in‑memory data structures (hash tables, skip lists) and a multiplexed I/O model. The event‑driven mechanism (select/poll/epoll) lets one thread monitor many sockets, avoiding blocking and enabling tens of thousands of operations per second.

Linux I/O Multiplexing Basics

File descriptors (FD) represent files, sockets, etc. epoll creates an epoll instance, registers FDs, waits for events, and processes ready events in a loop.

int epoll_create(int size);
int epoll_ctl(int epfd, int op, int fd, struct epoll_event *event);
int epoll_wait(int epfd, struct epoll_event *events, int maxevents, int timeout);
for (int i = 0; i < nfds; ++i) {
    if (events[i].events & EPOLLIN) {
        // handle read event
    }
    if (events[i].events & EPOLLOUT) {
        // handle write event
    }
}

Redis 6.0 Multithreaded I/O Model

When network I/O becomes the bottleneck, Redis can enable a pool of I/O threads. The main thread accepts connections, assigns sockets to I/O threads, which read and parse requests, while the main thread still executes the commands. After command execution, I/O threads write the responses back to the sockets.

Configuration (in redis.conf)

io-threads-do-reads yes
io-threads 6   # example for an 8‑core machine

Summary

Redis uses a single thread for command execution but adopts an event‑driven epoll model and optional multithreaded I/O (v6.0) to fully exploit multi‑core CPUs while keeping the core data path simple and fast.