Threading

Node.js operates on a single-thread event loop, using non-blocking I/O calls, allowing it to support tens of thousands of concurrent connections without incurring the cost of thread context switching.^[57] The design of sharing a single thread among all the requests that use the observer pattern is intended for building highly concurrent applications, where any function performing I/O must use a callback. To accommodate the single-threaded event loop, Node.js uses the libuv library—which, in turn, uses a fixed-sized thread pool that handles some of the non-blocking asynchronous I/O operations.^[7]

A thread pool handles the execution of parallel tasks in Node.js. The main thread function call posts tasks to the shared task queue, which threads in the thread pool pull and execute. Inherently non-blocking system functions such as networking translate to kernel-side non-blocking sockets, while inherently blocking system functions such as file I/O run in a blocking way on their own threads. When a thread in the thread pool completes a task, it informs the main thread of this, which in turn, wakes up and executes the registered callback.

A downside of this single-threaded approach is that Node.js does not allow vertical scaling by increasing the number of CPU cores of the machine it is running on without using an additional module, such as cluster,^[36] StrongLoop Process Manager,^[58] or pm2.^[59] However, developers can increase the default number of threads in the libuv thread pool. The server operating system (OS) is likely to distribute these threads across multiple cores.^[60] Another problem is that long-lasting computations and other CPU-bound tasks freeze the entire event-loop until completion.^{[citation needed]}