Proxy locations


North America

South America




See all locations

Network statusCareers

Back to blog

Concurrency vs Parallelism: The Main Differences

Iveta Vistorskyte

2021-07-016 min read

At first, it may seem that concurrency and parallelism may be referring to the same concepts. However, these terms are actually different.

This article explains the differences between concurrency vs parallelism. We also use a practical example to explore the concepts even more and show how using concurrency and parallelism can help speed up the web scraping process.

For your convenience, you can also watch this tutorial in a video format or keep reading a blog post.

What is concurrency?

To understand the difference between concurrency vs parallelism, let’s begin with a definition of concurrency. According to the Oxford Dictionary, concurrency means two or more things happening at the same time. However, this definition isn’t very helpful because parallel execution would also mean something similar. Let’s take a closer look at multitasking definition first. 

A Central Processing Unit (CPU, or simply a processor) can work on only one task at a time. If multiple tasks are given to it, e.g., playing a song and writing code, it simply switches between these tasks. This switching is so fast and seamless that, for a user, it feels like multitasking.

Concurrency is pausing and resuming threads

This capability of modern CPUs to pause and resume tasks so fast gives an illusion as if the tasks are running in parallel. However, this is not parallel. This is concurrent.

Concurrency can be broadly understood as multi-threading. There are usually many ways of creating concurrent applications, and threading is just one of them. Sometimes, other terms like asynchronous tasks are also used. The difference lies in the implementation and details. However, from a broader point of view, they both mean a set of instructions that can be paused and resumed. 

There is a programming paradigm called Concurrent Computing. This involves writing code so that it seems like more than one process is being performed simultaneously, while they never actually execute at the same time. This is known as concurrent programming

What is a thread?

In broad terms, a thread is the smallest set of tasks that can be handled and managed by the operating system without any dependencies on each other. The actual implementation differs in various operating systems. The way of programmatically creating threads also differs in various programming languages.

Python provides a powerful threading module for creating and managing threads. 

Practical example

To understand how concurrency works, let’s solve a practical problem. The issue is to process over 200 pages as fast as possible. Here are the details:

Step 1. Go to the Wikipedia page with a list of countries by population and get the links of all the 233 countries listed on this page.

Step 2. Go to all these 233 pages and save the HTML locally.

Let’s create a function to get all the links. At first, we won’t involve concurrency or parallelism here.

import requests
from bs4 import BeautifulSoup
from urllib.parse import urljoin
def get_links():
    countries_list = ''
    all_links = []
    response = requests.get(countries_list)
    soup = BeautifulSoup(response.text, "lxml")
    countries_el ='td .flagicon+ a')
    for link_el in countries_el:
        link = link_el.get("href")
        link = urljoin(countries_list, link)
    return all_links

This function gets the response from the link and uses BeautifulSoup to extract all the links. Links that we retrieve are relative links. They are converted to absolute links using urljoin

Now let’s write a function that doesn’t use any threading, but sequentially downloads the HTML from all those 233 links.

First, let’s create a function to fetch and save a link.

def fetch(link):
    response = requests.get(link)
    with open(link.split("/")[-1]+".html", "wb") as f:

This function is simply getting the response of the parameter link and saving it as an HTML file.

Finally, let’s call this function in a loop:

import time
if __name__ == '__main__':
    links = get_links()
    print(f"Total pages: {len(links)}")
    start_time = time.time()
    # This for loop will be optimized
    for link in links:
    duration = time.time() - start_time
    print(f"Downloaded {len(links)} links in {duration} seconds")

With our computer, this took 137.37 seconds. Our objective is to bring this time down.

Using concurrency to speed up processes

Although we can create threads manually, we’ll have to start them manually and call the join method on each thread so that the main program waits for all these threads to complete.

The better approach is to use the ThreadPoolExecutor class. This class is part of the concurrent.futures module. The benefit of using this class is that it allows us an easy interface for creating and executing threads. Let’s see how it can be used.

First, we need to import ThreadPoolExecutor:

from concurrent.futures import ThreadPoolExecutor

Now, the for loop written above can be changed to the following

with ThreadPoolExecutor(max_workers=16) as executor:, links)

Here, the executor applies the function fetch to every item of links and yields the results. The maximum number of threads is controlled by max_workers argument.

The final result is astonishing! All these 233 links were downloaded in 11.23 seconds. It’s a better result than the synchronous version which took around 138 seconds.

It’s important to find the sweet spot for the max_worker. On our computer, if the max_worker parameter is changed to 32, the time comes down to 4.6 seconds. Increasing this number further doesn’t improve things much. This sweet spot will differ for every processor for the same code.

Note: the print() function isn’t thread-safe. The reason is that print works with a reference to the standard output. The standard output is shared globally. We’ll use print() with a new line character explicitly and an empty end parameter.

What is parallelism?

In the previous section, we looked at a single processor. However, most processors have more than one core. In some cases, a machine can have more than one processor.

One example is parallel computing. This is a type of computation in which multiple processors carry out many processes at the same time. To achieve this parallel processing, specialized programming is needed. This is known as parallel programming, where the code is written to utilize multiple CPU Cores. In this case, more than one process is actually executed in parallel. The following image should help understand parallelism.

Parallelism is multiple threads running in multiple CPUs

Using parallelism to speed up processes

Let’s go back to the previously mentioned practical issue of downloading the HTML from all those 233 links.

In Python, parallelism can be achieved  by using multitasking. It allows us to download several links at the same time by using several processors.

To write an effective code that can be run on any machine, you would need to know the number of processors available on that machine. Python provides a very useful method, cpu_count(), to get the count of the processor on a machine. This is very helpful to find the exact number of tasks that can be processed in parallel. Note that in the case of a multi-core CPU, each core works as a different CPU.

Let’s start with importing the required module:

from multiprocessing import Pool, cpu_count

Now we can replace the for loop in the synchronous code with this code:

with Pool(cpu_count()) as p:, links)

This will create a multiprocessing pool that is equal to the count of the CPU. It means that the limit of multiple tasks being carried out would be determined when the code is actually running.

This fetches all 233 links in 18.10 seconds. It’s also noticeably faster than the synchronous version which took around 138 seconds.

Concurrency vs. parallelism: the differences

Here are the differences between concurrency and parallelism:

  • Concurrency is when multiple tasks can run in overlapping periods. It’s an illusion of multiple tasks running in parallel because of a very fast switching by the CPU. Two tasks can’t run at the same time in a single-core CPU. Parallelism is when tasks actually run in parallel in multiple CPUs.

  • Concurrency is about managing multiple instruction sequences at the same time, while parallelism is running multiple instruction sequences at the same time.

  • In Python, concurrency is achieved by using threading, while parallelism is achieved by using multitasking.

  • Concurrency needs only one CPU Core, while parallelism needs more than one. 

  • Concurrency is about interruptions, and parallelism is about isolation.

Combination of parallelism and concurrency

This is often known as Parallel Concurrent execution. The following image can help to understand the combination of parallelism and concurrency.

An example of concurrent and parallel execution

As it’s evident, if multiple CPUs are running multiple threads, the process is parallel and concurrent at the same time.

Suitable solution for web scraping

It can be assumed that both concurrency and parallelism make the web scraping process much faster. However, this should be taken with a pinch of salt. Every project is unique, and the complexity of every project is different. 

Starting with concurrency first and then looking at parallelism would be a great idea. Combining concurrency and parallelism may help in some cases, but it can make code complex and introduce challenging trace bugs. Once again, you should choose the one that suits your web scraping project best.

You can always forget about complex web scraping processes and choose an advanced public data collection solution – Web Scraper API. Try it for free to decide if it's a suitable option for your case.


In this article, we explored concurrency vs parallelism. Concurrency can be easily understood as threads or units of work that can be paused and resumed. Parallelism is simply multiple tasks running on multiple CPUs. Concurrency and parallelism can be achieved using Python threading and multitasking libraries, respectively. Using either concurrency or parallelism will improve the performance of the web scraping process significantly.

About the author

Iveta Vistorskyte

Lead Content Manager

Iveta Vistorskyte is a Lead Content Manager at Oxylabs. Growing up as a writer and a challenge seeker, she decided to welcome herself to the tech-side, and instantly became interested in this field. When she is not at work, you'll probably find her just chillin' while listening to her favorite music or playing board games with friends.

All information on Oxylabs Blog is provided on an "as is" basis and for informational purposes only. We make no representation and disclaim all liability with respect to your use of any information contained on Oxylabs Blog or any third-party websites that may be linked therein. Before engaging in scraping activities of any kind you should consult your legal advisors and carefully read the particular website's terms of service or receive a scraping license.

Related articles

Get the latest news from data gathering world

I’m interested