Concurrency vs Parallelism: A Practical Programming Explanation
Content edited . This editorial update does not represent a new runtime test.
On this page
Concurrency organizes multiple tasks so they can make progress during overlapping periods. Parallelism executes work at the same time. A single CPU core can interleave concurrent tasks; multiple cores can execute tasks in parallel.
A useful distinction
| Question | Concurrency | Parallelism |
|---|---|---|
| What does it describe? | How work is coordinated and progresses. | Whether work executes simultaneously. |
| Does it require several CPU cores? | No. | CPU work needs execution resources that can run simultaneously. |
| Can threads be involved? | Yes, but threads are not required. | Yes, but creating threads alone does not prove simultaneous execution. |
The concepts overlap: a concurrent program may use parallel execution, and a sequentially designed program may delegate a particular operation to a parallel implementation.
The illustration was included in the original 2018 article. Its precise original attribution has not been established in this revision, so the previous confident attribution has been removed. A primary conceptual reference is Rob Pike’s Concurrency is not parallelism.
Think about waiting and work
Imagine two requests. One waits for a database result while another can parse an already available response. An event loop can make progress on the second while the first waits, even with one execution thread. That is a useful concurrency arrangement.
Now consider two independent CPU-heavy calculations running on separate cores. Their computation can happen simultaneously: that is parallel execution. Synchronization, memory access and runtime constraints can still limit speedup.
Avoid common shortcuts
“Uses multiple threads” is not a complete definition of either concept. Nor does a concurrency mechanism guarantee better performance. Identify the workload, waiting time, shared state and correctness requirements before choosing an approach.
For service-capacity terminology, see web concurrency, QPS and latency. Concurrent requests and programming concurrency are related ideas at different levels; daily page views measure neither directly.