Skip to main content
HomeBlogData Engineering

What is Serverless Computing?

Learn about serverless computing, a cloud model where cloud providers manage infrastructure and allow software developers to ship applications quickly.
Jul 13, 2024  · 13 min read

Maintaining a physical or virtual server is no small feat, whether due to the high maintenance costs or the need for skilled manpower. Serverless computing solves these problems by providing a budget-friendly way of allowing developers to build and run applications on the cloud. 

In this article, we will review serverless computing, its applications, and its benefits to developers and businesses. This is an important topic because the market size for serverless computing exceeded $9 million in 2022 and is projected to expand another 25% in the next ten years. Let's jump right in.

Definition and Overview

Let’s say the light and water you use in your house are charged based on your usage instead of a fixed estimated monthly fee. This is how serverless computing works: It is just a way cloud providers allocate their infrastructure for you to build and run your applications based on usage without worrying about server maintenance.

Unlike traditional cloud computing, where a physical or virtualized server, storage, and networking equipment are set up, in serverless computing, the cloud provider manages the infrastructure and automatically allocates resources to you as your application scales. In other words, serverless computing is the abstraction of the server from developers, allowing them to focus more on the applications they are building instead of worrying about the infrastructure the application is hosted on. 

Imagine you are used to having 100 to 200 users daily on your application. If the next day your users suddenly increase to 1,000,000, the server will automatically scale your resources to meet that demand. This is not feasible with traditional cloud computing as the servers will experience downtime due to increased traffic, and if the server has low storage, larger storage has to be purchased, which takes time.

Category Serverless Computing Traditional Computing
Scaling Dynamic Fixed
Billing Based on usage Ongoing costs, regardless of usage, also include maintenance and operational costs
Infrastructure management Abstracted from businesses Requires active management

Key Characteristics

Before looking at key features that distinguish serverless computing from other traditional cloud models, let’s look at the key terminologies in serverless computing.

  • Function: This is a piece of code designed to perform a specific task, such as a file upload, processing HTTP requests, or something else. Each function is independent of each other and designed to act in response to an event.
  • Invocation: This is the calling of a function to perform a specific action, which you can also take as the execution of a function code in response to an event.
  • Duration: This is the time it takes for a function to execute from when it was invoked to when it returns a response.
  • Cold Start: This is the time taken to allocate resources by a serverless platform when a function is invoked before it starts execution.
  • Concurrency Limit: This is the maximum number of instances a function can run simultaneously in response to events and requests.
  • Timeout: This is the maximum duration allowed for a function before termination by the serverless platform.

Event-driven execution

Event-driven execution is also known as function-as-a-service (FAAS), where the serverless application is divided into independent serverless functions triggered by specific events from sources such as HTTP requests, changes in a database, message queries, or file uploads.

Suppose you have a serverless application that automatically processes images whenever they are uploaded to an Amazon S3 bucket. When a user uploads an image file, a function is triggered to process the image and save it in another Amazon S3 bucket. This ensures the function runs only when needed, making the system efficient and cost-effective. 

Image demonstrating event-driven execution, using file upload as an example.

Event-driven execution, using file upload as an example. Image by Author

Auto-scaling

Auto-scaling is the allocation of computing resources based on demand or increase in workloads. It’s a key characteristic of serverless computing that makes it efficient and adaptable, ensuring resources are not wasted if there is not much workload or demand on the server, and downtime is not experienced when the server experiences high demand.

For example, your application gets a spike in traffic. Instead of experiencing downtime on your server due to the high traffic, the platform automatically provides additional instances to handle the increased workload. Likewise, the platform scales down resources to minimize costs when the traffic is low. This ensures efficiency without manual intervention, making the application cost-effective and responsive in various scenarios. 

Pay-per-use billing

Serverless platforms charge users based on the resources used rather than the resources allocated. Unlike traditional cloud models where you might not use all of the allocated resources, in serverless platforms, you pay only for the compute time used. Costs are also metered, which means that you pay per invocation or duration of function execution, ensuring you are billed precisely for the resources you utilize. 

If, for example, your application processes 100 images from users in a given month, instead of paying for 24/7 server usage, you pay only for the compute time used for the 100 processed images. 

Server management abstraction

This is also a key feature of serverless computing where developers and businesses don’t have to worry about provisioning, scaling, and maintaining the servers. These allow them to focus on the core business problems and leave server maintenance to the cloud providers. 

Concurrency management

Serverless platforms can execute multiple functions simultaneously. This makes it fast and efficient when compared to traditional approaches. If users want to upload images on their platform and the serverless provider provides a default concurrency limit of 100, then any requests beyond the concurrency limit are queued and run in the next function execution. 

Benefits of Serverless Computing

Cost efficiency

Unlike traditional cloud computing, which requires dedicated servers and incurs costs whenever the server is idle, serverless platforms use an event-based model to charge developers. A useful analogy is using a taxi service instead of owning a car. With a taxi you don’t have to worry about the expenses spent on it such as parking fee, fuel and so on, you get to pay only for the distance you travel. In serverless computing, just like the taxi, you only pay for the computing resources you use. Unlike the traditional approach (owning a car), you are responsible for every computing resource you use, even when the system is idle. 

Image showing cost efficiency comparison between traditional servers and serverless computing.

Cost efficiency comparison between server types. Source: Cloudflare

Reduced operational complexity

Much time is saved in serverless computing because developers don’t need to devote their time to installing and maintaining servers; they focus more on building the application. Since applications are built as independent cloud functions, one can update a function without disrupting others or the entire application.

Improved scalability

This is one of the most important benefits of a serverless platform and why it is popular among smaller organizations and startups. Serverless platforms make it easy for developers to scale their operations automatically when demand increases. For functions that experience request fluctuations, serverless platforms scale to meet these requests by increasing or decreasing resource allocation, hence ensuring computing resources are optimized. 

Faster time to market

With serverless applications, you can release applications quickly and immediately get user feedback. This is important for startups as this reduces the time spent and the manpower needed to build applications.

Reliability

Unlike applications hosted on dedicated servers, one can run serverless applications from anywhere. This improves application performance and reduces latency compared to traditional cloud computing.

Serverless Architecture and How It Works

Serverless architecture is a way of building applications without installing and managing the infrastructure hosting the application. Developers can build serverless applications using either of the two serverless models; either using Backend as a Service (BaaS) or Function as a Service (FaaS).

  • Backend as a Service (BaaS): This serverless model is usually used when building web and mobile application backends. Developers don’t need to code backend features like databases, user authentication, storage, and others. This makes development fast as developers mainly focus on the application frontend. Examples of providers offering BaaS, are Firebase, SupaBase, and AWS Amplify.
  • Function as a Service (FaaS): In FaaS, developers write the code for the platform to execute without worrying about managing computing resources and scaling. This allows the execution of functions in response to triggered events. Examples of providers offering FaaS are AWS Lambda, Azure Cloud Functions, and Google Cloud Functions.

Image showing FAAS and BAAS serverless cloud architectures

FAAS and BAAS serverless cloud architectures. Source: Journal of Cloud Computing

The main aim of serverless architecture is to abstract server management from developers. Here is a breakdown of how serverless platforms operate in their infrastructure.

  • Function Creation: Applications built using a serverless architecture have their code written into subparts, where each part is an individual function independent of other functions and meant to achieve a specific task. 
  • Function Development: These functions are packaged and deployed to a serverless platform like AWS Lambda, Azure Cloud Functions, or Google Cloud Functions.
  • Event-Driven Execution: These functions are invoked in response to a particular event or trigger. These triggers are either database changes, HTTP requests, or something else. 
  • Auto-Scaling: The serverless platforms handle the use of resources by events as workload increases or decreases. For example, the server allocates more resources when a particular function receives high traffic.
  • Transient Containers: Triggering an event creates containers that allow the executed function to access required resources. After completion of a task, these containers are then destroyed.
  • Billing: Billing is done based on the amount of execution time and compute resources used. 
  • Statelessness: Serverless functions, when executed, don’t retain information. any needed information in an invocation is saved either in a database or externally. 
  • Logs and Monitoring: Serverless platforms provide logging and monitoring tools to track application performance and issues in functions.

One might wonder: Isn’t serverless architecture the same as container architecture since both abstract the server from the developers? This is true, but unlike serverless functions, which completely abstract the server when building and shipping applications in containers when an application experiences high traffic, one needs to scale the containers using tools like Kubernetes. This defeats the aim of FaaS, where all actions related to the server are handled automatically by the serverless platform.

In a container architecture, you can have container instances that can run for a long period which can incur costs, unlike in serverless functions where you are billed for the amount of time your function spends when running. For smaller applications, using a serverless architecture makes it easy to decouple it into parts that the serverless platforms can run as independent functions.

Serverless Computing Platforms

Before the advent of Google App Engine in 2008, Zimki offered the first “pay as you go” platform for code execution, but was later shut down. In its early stage, the Google App Engine could only support Python and featured metered billing for applications which among them was SnapChat. Around 2010, another platform called PiCloud also provided FaaS support for Python applications.

In 2014, AWS popularized the serverless model releasing tools such as AWS Serverless Application Model (AWS SAM), and Amazon CloudWatch. Google then released their second serverless offering Google Cloud functions alongside Azure functions in 2016. Since then various serverless platforms have been released such as Function Compute by Ali Baba Cloud and IBM Cloud Functions by IBM Cloud.

To eliminate the need for a virtualized or physical database, serverless databases have also been developed. AWS offers Amazon Aurora, a serverless version based on MySQL and PostgreSQL. Azure offers Azure Data Lake, and Google provides Firestore.

Serverless cloud platforms. Source: Network Interview

Practical Applications and Use Cases

Websites and APIs

Serverless computing is used to build website applications and REST APIs. The interesting part is that applications are built using a serverless infrastructure auto-scale based on user demands, improving user experience.

Media processing

Serverless architecture makes it easier to handle media processing. Users can upload media content from varying devices and sizes and the platform processes a single function that meets the demand of every user without reducing application performance. For example, a user can upload an image via an S3 bucket which triggers an AWS Lambda function to either add a watermark or thumbnail to the image.

Chatbots

Developers can implement a chatbot to respond to customer questions using serverless architecture and pay only for the resources the Chatbot uses. For example, Slack uses serverless architecture to handle the varying requests of bots to avoid underutilized bandwidth due to the everyday fluctuation of customer needs.

Webhooks

You can use a serverless platform to interact with SaaS vendors through a webhook HTTP endpoint, which receives notifications and performs tasks. This offers minimal maintenance, low costs, and automatic scaling for the built webhook.

IoT

Coca-Cola uses serverless architectures on their vending machine Freestyle to allow customers to order, pay, and receive payment notifications for their beverages. Coca-Cola claimed they were spending around $13,000/year to operate their vending machines, which was reduced to $4,500/year after implementing serverless architectures on the vending machines. 

Data processing

Major League Baseball Advanced Media built their product Statcast with serverless architecture to provide users with accurate and real-time sports metrics. It uses serverless computing to process data and give users insights about baseball games. 

Event-driven applications

Serverless architecture is used for event-driven applications, such that when an event or a state is changed, it triggers a service. One can use serverless platforms to watch for changes to a database and compare the changes against quality standards.

Serverless Edge Computing

Due to the long-distance travel of data, applications take time to process requests and deliver content from centralized servers, which introduces latency and bottleneck issues.

Serverless edge computing solves this issue by distributing computing resources to multiple locations to reduce the workload on the central server. Serverless edge computing is the location of computing resources that run serverless functions closer to end users (edge). This allows the serverless application to operate on more devices, decrease congestion, and lower latency.

In serverless edge computing, infrastructure is dedicated to each device, allowing it to perform complex tasks without sending data back to the central location for processing. Here are some use cases that optimize user experience using serverless edge functions:

  • Personalized Experience: With serverless edge functions, users can personalize content based on preferences, location, and device type. 
  • Video Streaming and Gaming: Processing requests close to users reduces the latency and buffering experienced by streaming video or playing online games.
  • Security and Authentication: By distributing workload to various edge locations, malicious traffic is checked and filtered before it reaches the central infrastructure.
  • IOT Devices: Serverless edge computing makes it easy for IOT devices to perform efficiently by ensuring that the infrastructure running the device is close to the users. 

Challenges and Considerations

Despite the benefits brought by serverless computing, it faces some of the following drawbacks.

  • Vendor Lock-in: Most cloud providers provide several services that integrate well with serverless applications. Although it’s possible to use another vendor’s service in your server application, services provided by a single provider are easy to integrate. 
  • Less Control: Another issue is the lack of control over the server. If there is a problem with the server, such as an outage or hardware problem, you can’t fix it and are left with the cloud provider to fix it.
  • Cold Starts: Cold start occurs when a function that has been inactive for a while is invoked and it takes several seconds to execute. This can introduce latency which can affect application performance and also user experience.
  • Security: By entrusting your data to a third party and sharing a server with other users, you risk exposing your application and user data if the servers are not configured correctly.
  • Debugging Complexities: It’s difficult for developers to replicate the production environment locally due to the features of serverless applications, hence making it difficult to test the code locally. Integration tests that test the interaction between frontend and backend code are also difficult to perform when in a serverless environment. 

Conclusion

For companies interested in building lightweight applications, serverless computing is the way forward. For applications made up of a large number of services with complex interactions, a hybrid infrastructure made up of virtual machines for large processes is recommended; in this case, the serverless containers are used only for short tasks. Serverless edge functions can also be considered because they minimize latency by processing data locally without putting much workload on the central server.

Keep in mind that various advancements and developments are still being made to expand the capabilities of serverless computing. For example, a multi-cloud approach can be used to build serverless applications, where one can build a serverless application using services from more than one cloud provider. More developments have also been made to ensure zero cold start of serverless functions, as implemented in Cloudflare Workers

Thanks for tuning in. To learn more about serverless computing and build your skills in the field, check out the following DataCamp resources:

For shorter reading, check out our blog post for project ideas for all skill levels.


Photo of Adejumo Ridwan Suleiman
Author
Adejumo Ridwan Suleiman
LinkedIn

Experienced data science instructor and Biostatistician with expertise in Python, R, and machine learning.

Frequently Asked Questions

What is serverless computing?

Serverless computing is a cloud computing model where cloud providers manage the infrastructure, automatically allocating resources as applications scale. Developers can focus on building and running applications without worrying about server maintenance.

How does serverless computing differ from traditional cloud computing?

Unlike traditional cloud computing, serverless computing offers dynamic scaling, pay-per-use billing, and abstracts infrastructure management from businesses.

What are the main benefits of serverless computing?

Key benefits include cost efficiency, reduced operational complexity, improved scalability, faster time to market, and increased reliability.

What are some common use cases for serverless computing?

Serverless computing is used for building websites and APIs, media processing, chatbots, webhooks, IoT applications, data processing, and event-driven applications.

What challenges should be considered when adopting serverless computing?

Some challenges include vendor lock-in, less control over the underlying infrastructure, cold starts, potential security concerns from sharing resources, and debugging complexities due to the distributed nature of serverless applications.

Topics

Learn with DataCamp

Course

Understanding Cloud Computing

2 hr
84.2K
A non-coding introduction to cloud computing, covering key concepts, terminology, and tools.
See DetailsRight Arrow
Start Course
See MoreRight Arrow
Related

blog

Cloud Computing and Architecture for Data Scientists

Discover how data scientists use the cloud to deploy data science solutions to production or to expand computing power.
Alex Castrounis's photo

Alex Castrounis

13 min

blog

Private Cloud vs. Public Cloud: Which One to Choose?

Cloud computing offers public, private, and hybrid models, each with distinct advantages and tradeoffs, requiring careful consideration to select the right fit for your organization.
Maria Eugenia Inzaugarat's photo

Maria Eugenia Inzaugarat

9 min

blog

Top 20 AWS Lambda Interview Questions and Answers for 2024

AWS Lambda is a serverless compute service and an increasingly common subject in technical interviews. Whether you're new to cloud computing or a seasoned pro, understanding AWS Lambda is essential.
Zoumana Keita 's photo

Zoumana Keita

12 min

blog

What is Google Cloud Platform (GCP)? A Comprehensive Guide to Mastering Cloud Services

Learn what Google Cloud Platform is, from cloud basics to advanced analytics and AI. Become an expert in GCP core features and strategic advantages.
Jana Barth's photo

Jana Barth

16 min

tutorial

Snowflake Tutorial For Beginners: From Architecture to Running Databases

Learn the fundamentals of cloud data warehouse management using Snowflake. Snowflake is a cloud-based platform that offers significant benefits for companies wanting to extract as much insight from their data as quickly and efficiently as possible.
Bex Tuychiev's photo

Bex Tuychiev

12 min

tutorial

AWS EC2 Tutorial For Beginners

Discover why you should use Amazon Web Services Elastic Compute Cloud (EC2) and how you can set up a basic data science environment on a Windows instance.
DataCamp Team's photo

DataCamp Team

7 min

See MoreSee More