In cloud computing, commercial providers make computing resources available on demand to their customers over the Internet. This service is partly offered "serverless", that is, without servers. How can that work? Computing resources without a server, isn’t that like a restaurant without a kitchen?
"The term is misleading," says computer science Professor Samuel Kounev from Julius-Maximilians-Universität (JMU) Würzburg in Bavaria, Germany. Because even serverless cloud A recent article in the magazine of the Association for Computing Machinery (ACM) deals with the history, status and potential of serverless computing. Among the authors are Samuel Kounev and Dr. Nikolas Herbst, who heads the JMU research group "Data Analytics Clouds".
Experts Define Serverless Computing InconsistentlyThe origins of the research article lie two years back. In 2021, around 50 international experts in serverless computing met at a seminar at Schloss Dagstuhl - Leibniz Centre for Informatics.
"We discussed the most important developments and research questions there and found out that many of us define serverless computing differently and that contradictions sometimes arise," Kounev recounts. This is not unusual in science, he says, when a paradigm shift emerges.
In order to create clarity, a small group of researchers got together. Together, they tried to define serverless computing precisely after the seminar. Exactly this circle of people has now published the article in the renowned CACM magazine; it also includes Ian Foster, a prominent pioneer of cloud computing from the USA.
Two Key Principles DefinedThe team has defined two principles that characterise serverless computing. "NoOps" is the first, which stands for "no operations". This means, as described above, that the technical server management, including the hardware and software layers, is completely in the responsibility of the cloud provider.
The second principle is "utilisation-based billing", which means that only the time during which the customer actively uses the allocated computing resources is billed. With classic cloud services, on the other hand, a fee is paid for the entire time during which the respective cloud resources were rented.
Start image of a video in which Würzburg computer science Samuel Kounev explains what ’serverless computing’ is - and what it is not. (Image: Screenshot von Vimeo.de / ACM)
At JMU, the computer scientists in Samuel Kounev’s team are working, for example, on the elasticity of cloud services, especially the autoscaling of computing resources. The goal here is that the amount of computing resources allocated to a cloud application automatically adapt to increasing or decreasing demand over time.
A long-term project that Kounev’s team wants to realise in the coming years is a serverless cloud platform for large workflows in Earth observation. The focus is on climate research with satellite data. Scientists should be able to use the platform to study various global effects of climate change easily, quickly and efficiently. These include changes in forests, snow cover in mountains, or biodiversity.
In this field, Kounev’s team is cooperating with the JMU Chair of Remote Sensing, the German Aerospace Center (DLR), the Leibniz Computing Centre of the Bavarian Academy of Sciences and Humanities and the Max Planck Institute for Behavioural Biology (Radolfzell / Konstanz), among others.
PublicationServerless Computing: What It Is, and What It Is Not? Samuel Kounev; Nikolas Herbst; Cristina L. Abad; Alexandru Iosup; Ian Foster; Prashant Shenoy; Omer Rana; Andrew A. Chien in Commun. ACM (2023). 66(9) 80-92. DOI: 10.1145/3587249, published online 23 August 2023.
ACM has produced a with Professor Samuel Kounev to accompany the publication.
By Robert Emmerich