In this article you will read

• what serverless computing means,  
• what advantages serverless computing has over containers and  
• which service you can use to work “serverless” on the Open Telekom Cloud.

It's not just the cars on our roads that are increasingly “software-defined”. Many functions in a wide range of end devices are implemented using software. This software is either made available via the web or is permanently integrated into the devices. “Software makes the world go round”, one might say, or as Marc Andreessen put it: “Software is eating the world.”

In a world in which more and more processes are controlled by software, the rapid provision of software-defined services is becoming a supreme discipline for companies. Many current IT trends are aimed at relieving developers and making software development faster and more efficient. These include agile development methods and the recycling of successful code. On the other hand, DevOps means that software developers are increasingly forced to look beyond the development stage towards operations. This in turn increases the workload.

The cloud has its merits when it comes to developing software in a modern way and making it available quickly. It is not without reason that it first established itself in the developer environment. Complete development and test environments provided within minutes which are tailored to the needs of the developer, or project management for collaborative development from the cloud are commonplace today.

Development from the cloud is one side of the coin, operating applications is the other. As elegant as the use of the cloud is: Developers must always take a look at infrastructures. What resources are required for optimal (i.e. high-performance and cost-efficient) operation? Wouldn't it be a huge advantage to end this discussion once and for all? Resources that are automatically provided appropriately without the need to select, assign, configure etc. infrastructure components?

“Serverless” or Function as a Service (FaaS) is the answer to this question. With the help of FaaS or "serverless", developers (and later users) access ready-to-use, complete services. They can select these from libraries and add them to their application. These ("all inclusive") services independently take care of everything they need for optimal operation – including the appropriate infrastructure.

Abstraction of infrastructure resources – this desire is as old as the cloud itself. Serverless computing is also not the only solution for moving away from manual provisioning or clicking on a management console. Containers have often been used for this purpose. However, containers and serverless computing are two different approaches to providing and operating applications in the cloud. Both have their advantages and disadvantages. Here are the differences:

• When using containers, container images are created, deployed, and managed on a container orchestration platform such as Kubernetes or Docker. Serverless computing eliminates this workload.
• Anyone using containers must take care of managing scaling at cluster level. Serverless computing automates scaling at functional level (“all inclusive“).
• Containers are ideal for continuously running applications and processes, while “serverless“ functions are generally better suited to short, event-driven tasks with limited execution times (“ad hoc”).
• Containers are usually paid for according to the allocated resources, regardless of whether they are used or not. Serverless computing is paid according to the actual requests and the duration of execution.

Serverless functions can theoretically take on tasks of any complexity. As a rule, they are provided for standard tasks in libraries such as GitHub or in software development kits. In addition, developers can add their own functions to the library at any time.

A typical example of a possible function: an application is operated in the cloud. During use, users must upload images to an object storage. They should receive a preview with thumbnails so that they can check whether they have uploaded the correct images. Instead of developing the service itself, the application developer uses a function from a library. The function is triggered when images are uploaded (“trigger”). It is linked to the object storage and displays the thumbnails in real time. Another function could also be linked to the upload trigger: The second function holds an AI-based service for image recognition, which automatically adds keywords to the images. The function is terminated immediately after the upload. All associated resources are switched off.

With FunctionGraph, the Open Telekom Cloud also offers the option of using serverless functions “as a service”. Functions such as Kafka or GaussDB are provided as standard in the library.  

But FunctionGraph primarily offers developers the opportunity to run their own code without needing to provide or manage servers. Developers only need to upload their code and set conditions for its execution. FunctionGraph takes care of the rest.

• Write code in Node.js, Python, Java, C#, PHP or Go
• Upload code as a .zip or .jar file. Developers have the option of editing the code inline
• Set the trigger functions either as an API call or as a cloud service event
• Auto scaling: FunctionGraph scales automatically based on the requests to the function without requiring any configurations
• Monitoring: Run logs of the functions are collected in the Log Tank Service
• The logs can be viewed and analyzed by the developer via the Application Operations Management service
• Billing is based on the number and duration of calls to the function

By providing their own function within a library, developers not only gain the opportunity to develop faster and more conveniently themselves, but also to offer their own functions as services on the market. FunctionGraph thus paves the way for new business models.