Redis: The single-thread REACTOR Model

Pay attention to the public account, communicate together, wechat search: sneak forward

Pure memory access, all data is in memory, all operations are memory-level operations, and memory response time is nanosecond. Therefore, the CPU of redis process has no disk I/O latency and memory read/write performance problems. CPU is not the bottleneck of Redis (memory size and network I/O are the bottleneck of Redis, namely the network transfer delay between client and server).
Using single thread model, single thread implementation is simple. Avoids the overhead of multiple threads with frequent context switches and synchronization mechanisms such as locking
Simple and efficient basic data structures: dynamic string (SDS), linked list, dictionary, skip linked list, integer set and compressed list. Redis then builds on this foundation to implement objects that users can manipulate: strings, lists, hashes, collections, ordered collections, etc
A network event handler for the REACTOR schema. It uses I/O multiplexing to monitor multiple sockets simultaneously, which is an efficient I/O model. Take a look at the IO+ REACTOR model for Linux high-performance networks

Single thread is used to avoid unnecessary context switching and race conditions. There are no cpu-consuming switches caused by multithreading
There are no locks to worry about, no locking and unlocking operations, and no performance costs due to possible deadlocks
Simple and maintainable, multithreaded mode will make the writing of the program more complex and troublesome, single thread implementation is easy to implement

We often refer to REDIS Single threaded as its Network event processing model REACTOR is single threaded
If a client request deletes millions of keys, this command can cause a block. In Redis 4.0, multithreading is chosen to implement this type of non-blocking command
In Redis6.0, multi-threading was officially introduced to handle client I/O read and write responses, while redis command events are still handled in the main thread
If it is for the entire Redis system, there are always other threads handling asynchronous tasks, such as AOF, RDB synchronization

Redis is fast not only because it is a single-threaded pure memory system, but also because it uses the Reactor model to process external requests using I/O multiplexing, reducing network connection and read/write latency. In this way, it can concurrently process a large number of client requests in network I/O operations, achieving high throughput and high concurrency