Concurrency vs Parallelism¶

Concurrency¶

Concurrency is the ability to make progress on multiple tasks during the same period of time.

The tasks do not have to execute simultaneously.

Instead, the system switches between tasks whenever one task is waiting, allowing multiple tasks to progress together.

Imagine a backend server handling multiple API requests.

Request 1 is waiting for PostgreSQL.
While it's waiting, the CPU starts processing Request 2.
When Request 2 waits for Redis, the CPU starts Request 3.

Even though only one request may be executing at a particular instant, all requests continue making progress.

Conceptually,

Time ───────────────────────────────────────────►

Request 1   ███ Waiting ███ Running ███ Done

Request 2        ███ Running ███ Waiting ███ Done

Request 3             ███ Running ███ Waiting ███ Done

The CPU stays busy by switching between tasks instead of waiting.

Parallelism¶

Parallelism is the ability to execute multiple tasks at exactly the same time.

This requires multiple CPU cores or processors.

For example, on a machine with four CPU cores,

Core 1 → Request 1

Core 2 → Request 2

Core 3 → Request 3

Core 4 → Request 4

All four requests execute simultaneously.

No context switching is required because each task has its own CPU core.

Concurrency vs Parallelism¶

Concurrency

        CPU

Task A ──────► Waiting
                 ▲
                 │
Task B ──────────┘
        │
        ▼
Task C

One CPU switches between multiple tasks.

Parallelism

Core 1 ───► Task A

Core 2 ───► Task B

Core 3 ───► Task C

Core 4 ───► Task D

Multiple CPU cores execute tasks simultaneously.

Backend Example¶

Suppose your application receives 100 API requests.

Each request performs the following operations.

Read from PostgreSQL
Read from Redis
Call an external payment service

Most of the time is spent waiting for I/O.

Without Concurrency¶

Request 1 -----------------------> Done

Request 2 -----------------------> Done

Request 3 -----------------------> Done

Every request waits for the previous one to finish.

With Concurrency¶

Request 1  Waiting PostgreSQL

Request 2  Running

Request 3  Waiting Redis

Request 4  Running

Request 5  Waiting Payment API

Whenever one request waits for I/O, another request gets CPU time.

This improves resource utilization and increases throughput.

With Parallelism¶

CPU Core 1 → Request 1

CPU Core 2 → Request 2

CPU Core 3 → Request 3

CPU Core 4 → Request 4

Multiple requests execute at the same time.

This is ideal for CPU-intensive workloads.

When Should You Use Each?¶

Concurrency¶

Best for I/O-bound tasks.

Examples:

Web servers
Database queries
Network requests
File I/O
External APIs

Parallelism¶

Best for CPU-bound tasks.

Examples:

Image processing
Video encoding
Machine learning
Mathematical computations
Large data processing

Relationship Between Them¶

Concurrency and parallelism are related but different concepts.

               Multiple Tasks
                     │
        ┌────────────┴────────────┐
        │                         │
        ▼                         ▼
Concurrency                Parallelism

Makes progress             Executes tasks
on many tasks              simultaneously

May use                    Requires
one CPU                    multiple CPU cores

A concurrent program is not necessarily parallel.

A parallel program is usually concurrent because multiple tasks are progressing simultaneously.

Follow-up Questions¶

What is concurrency?¶

Concurrency is the ability to make progress on multiple tasks during the same period of time.

The tasks may execute by taking turns.

What is parallelism?¶

Parallelism is the ability to execute multiple tasks simultaneously using multiple CPU cores.

Does concurrency require multiple CPU cores?¶

No.

A single CPU can achieve concurrency by switching between tasks.

Does parallelism require multiple CPU cores?¶

Yes.

True parallel execution requires multiple CPU cores or processors.

Which is better for web servers?¶

Concurrency.

Web servers spend most of their time waiting for databases, APIs, or network responses.

Which is better for CPU-intensive work?¶

Parallelism.

CPU-heavy tasks benefit from multiple CPU cores.

Common Interview Mistakes¶

❌ Thinking concurrency means tasks execute simultaneously.

✔ Concurrency means multiple tasks make progress together.

❌ Thinking concurrency requires multiple CPU cores.

✔ A single CPU can achieve concurrency through task switching.

❌ Thinking parallelism is always faster.

✔ Parallelism helps only when the workload can effectively use multiple CPU cores.

Interview Tip¶

A simple way to explain the difference is:

Concurrency is about dealing with multiple tasks at once.

Parallelism is about executing multiple tasks at once.

This distinction is easy to remember and is widely accepted in interviews.

Key Takeaways¶

Concurrency allows multiple tasks to make progress together.
Parallelism executes multiple tasks simultaneously.
Concurrency does not require multiple CPU cores.
Parallelism requires multiple CPU cores.
Concurrency is best suited for I/O-bound workloads.
Parallelism is best suited for CPU-bound workloads.
In Python, threads and asyncio provide concurrency, while multiprocessing provides true parallelism.