Intelligent Serverless and Cloud Applications Symposium

June 17-18, 2024, Zurich, Switzerland

Zurich University of Applied Sciences, Computer Science Teaching Building, Lagerstrasse 41, Zurich

Supported by:

Swiss National Science Foundation, CLOUDSTARS Cloud Open Source Research Mobility Network, Zürich Tourism, Zurich University of Applied Sciences

Symposium Programme [final]

Monday, June 17

15.00Opening and IntroductionJosef Spillner
15:30Talk session 1: Scaling and orchestration
A Dynamic Kubernetes Load Generation Solution Mimicking Human Traffic Patterns (#)
Autoscaling Rethought: A Continuous, Decentralized Approach
Towards Elastic Data Infrastructures

Ranjan Ojha
Martin Sträßer
Pedro García López
16:30Invited talk: Decentralized and Stateful Serverless Computing on the Internet Computer Blockchain (#)Alexandru Uta
17.30Community viewpointsVarious

Tuesday, June 18

9:00Pitch session: Advanced Big Data ApplicationsVarious
10:30Talk session 2: Security and reliability
Cost of gang-scheduling
Serverless Confidential Containers: Challenges and Opportunities (#)
Code Portable Serverless Workflows with Interoperable Cloud Services in Sky Computing (#)

Jawad Tahir
Carlos Segarra
Sashko Ristov
11:30Cloudstars demos/posters sessionVarious
12:00Talk session 3: Infrastructure engineering
Understanding Datacenter Scheduler Programming Abstractions
Case Studies Cloud Performance Prediction & Infrastructure as Code
Understanding Go compilation to WebAssembly bytecode (#)

Sacheendra Talluri
Jürgen Cito
Safia Guellil
14:00Invited talk: High Performance Serverless for HPC and CloudMarcin Copik
14:30Talk session 4: Serverless applications
Addressing General Data Protection Issues in Bioinformatics & ML Data (#)
Understanding Serverless Workloads through the study of Serverless Digital Mobility and other Novel Applications (#)
Microcredentials on Blockchain

Tuğba Önal-Süzek
Cristina Abad
Oliver Cvetkovski & Cihan Sakman
15:30Community paper discussion and draftingVarious


A Dynamic Kubernetes Load Generation Solution Mimicking Human Traffic Patterns

Abstract: Autoscaling is one of the most advertised feature of Kubernetes. While much research has gone in the subject, it can be observed that the only Horizontal Pod Auto Scalers have seen popularity among people. HPA while simple unfortunately have quite a few limitations which doesn't allow people to really take the full advantage of their resources. This lack of diversity is also because there isn't a reliable framework for people to compare the various autoscalers for their usecases. With our current work we aim to provide a reliable framework for comparing various scalers for their usecases so that they can reliably choose the autoscaler that best fits their needs. It does so by reading in the current behavior of the application as measured externally, then simulating the application in a sandbox environment. This also allows us to study the application behavior in the long term serving as a basis for developing a new autoscaler that is able to autoscale not only in cluster but also across different clusters.

Bio: Ranjan Ojha is currently working as a Research Assistant in the research group of Distributed Systems in Zurich University of Applied Sciences. His current focuses are on enhancing autoscaling in kubernetes as well as handling big data in a distributed system.

Decentralized and Stateful Serverless Computing on the Internet Computer Blockchain

Abstract: The Internet Computer (IC) is a fast and efficient decentralized blockchain-based platform for the execution of generalpurpose applications in the form of smart contracts. In other words, the IC service is the antithesis of current serverless computing. Instead of ephemeral, stateless functions operated by a single entity, the IC offers decentralized stateful serverless computation over untrusted, independent datacenters. Developers deploy stateful canisters that serve calls either to end-users or other canisters. The IC programming model is similar to serverless clouds, with applications written in modern languages such as Rust or Python, yet simpler: state is maintained automatically, without developer intervention. In this paper, we identify and address significant systems challenges to enable efficient decentralized stateful serverless computation: scalability, stateful execution through orthogonal persistence, and deterministic scheduling. We describe the design of the IC and characterize its operational data gathered over the past 1.5 years, and its performance.

Bio: Alex enjoys working on large-scale systems and infrastructure with a keen eye on designing and building efficient and performant systems. He has experience in both industry and academia having worked as an Assistant Professor at Leiden University in the Netherlands and as a systems researcher at AWS Lambda. His current goal is to help make the Internet Computer fast and reliable.

Serverless Confidential Containers: Challenges and Opportunities

Abstract: Serverless computing allows users to execute pieces of code (so called functions) on-demand in the cloud without having to provi- sion any hardware resources. However, by executing in the cloud and delegating control over hardware resources, the integrity of the execution and the confidentiality of function code and data are at the mercy of the cloud provider and serverless runtime. Confi- dential computing aims to remove trust from the cloud provider by executing applications inside hardware enclaves. In spite of the increasing adoption of confidential computing, designing a confi- dential serverless runtime with moderate performance overhead remains an open challenge. In this talk we present our experience porting the Knative serverless runtime to a confidential setting using Confidential Containers (CoCo), a technology that allows the execution of unmodified (encrypted) container images inside confidential VMs (cVMs). Our results show that cVMs are not ready to execute container-based serverless functions. Starting a serverless function in a CoCo from an encrypted container image with attestation takes up to 17 seconds. Starting 16 serverless functions concurrently takes more than three minutes, 20× slower than its non-confidential counterpart. We analyze the main sources of overhead, and outline the research challenges to bridge the gap between confidential and serverless computing.

Bio: Carlos is a fourth year PhD student at the Large-Scale Data and Systems Group (LSDS) of the Imperial College London, under the supervision of Prof. Peter Pietzuch. His research addresses the design and implementation of secure, high-performance, runtimes for the cloud. Carlos is particularly interested in privacy-preserving systems and confidential computing. All of his work is open-source and available on Github. During his PhD, Carlos has interned at Intel Labs and IBM Research.

Code Portable Serverless Workflows with Interoperable Cloud Services in Sky Computing

Abstract: Cloud providers introduce SDKs to enable easy access to their services. Unfortunately, these SDKs lock the users into cloud provider's environment. This vendor lock-in approach restricts the users to freely select individual services of multiple providers. This talk will present our advances in coding serverless workflow applications that are applicable for Sky computing. The main goal is to develop portable serverless workflows that can dynamically and transparently select interoperable cloud services across multiple cloud providers.

Bio: Dr. Sashko Ristov is an Assistant Professor for computer science at the University of Innsbruck, Austria. His main research interests include performance modeling and optimization of distributed systems and applications. The current focus is on serverless computing, cloud engineering, and cloud federation. In particular, he works on designing and building platforms that can support the entire life-cycle of distributed applications in federated cloud environments, including development, deployment, modeling, orchestration, optimization, scheduling, and execution.

Understanding Go compilation to WebAssembly bytecode

Abstract: WebAssembly (Wasm) is a relatively recent low-level language that is rapidly gaining popularity for enabling high-performance executions on the web. However, the design of this language is not just for the web; thanks to the WebAssembly System Interface (WASI), it can also run in other environments such as cloud, edge or embedded. Although Wasm has been well analyzed in languages with "unmanaged" memory, there is a gap in research in "managed" languages like Java and Golang, which employs a garbage collection (GC) algorithm. Golang, popular for developing microservices, has a TinyGo compiler, tested on edge devices and constrained environments which is able to compile to WebAssembly. This presentation explores the implications of cross-compiling Go code to WebAssembly using TinyGo, with a focus on garbage collection. We discuss factors influencing GC performance and how to control its impact on speed, memory usage, and concurrency.

Bio: My name is Safia Guellil. I am a PhD student and computer engineer at Rovira i Virgili University (URV) in Catalonia, Spain. I hold a Master's degree in Security Engineering and Artificial Intelligence from the same institution and currently, I am part of the Cloud and Distributed Systems Lab research group (URV). My research line focuses on enhancing Edge Computing with WebAssembly, a portable intermediate binary format for programming languages.

Addressing General Data Protection Issues in Bioinformatics & ML Data

Abstract: This talk will explore the ethical and data protection challenges in cloud storage and processing of sensitive biometric data for machine learning applications. At the context of our recent work of an orthodontics pre-diagnosis application, I will discuss general methodologies for secure data collection, GDPR compliance, and data security, emphasizing anonymization, encryption, and controlled access. The importance of informed consent and transparent communication with patients will be highlighted, alongside the role of AI and the necessity for human oversight in medical applications. Additionally, I will outline our current protocols for managing incidental findings to ensure ethical and professional handling of unexpected results to protect sensitive biometric data.

Bio: Dr. Tuğba Önal Süzek is the Head of the Bioinformatics Master's and Ph.D. Programs at Muğla University and the CEO of Kedi Mobile Applications, Inc., a start-up company. She earned her B.Sc. in Computer Engineering from Middle East Technical University, her M.S. in Computer Science from Johns Hopkins University, and her Ph.D. in Bioinformatics from George Mason University. Before returning to Turkey, she worked as a Staff Scientist at the NIH for 12 years, where she implemented and managed the SAGE, GEO, and PubChem BioAssay databases and software pipelines for transcriptomics and chemoinformatics data. Currently, Dr. Süzek is the principal investigator or co-investigator of two EU-funded projects and several national grants supporting research in bioinformatics and machine learning for medical applications. Her current research interests are in cancer genomics, machine learning of biomedical data and chemoinformatics. She has 19 international SCI publications with a Web of Science Index of 12.

Understanding Serverless Workloads through the study of Serverless Digital Mobility and other Novel Applications

Abstract: Serverless platforms already serve workloads that are significantly more diverse than originally envisioned. However, these frameworks should be continuously improved to better serve emerging applications. To understand how to improve serverless frameworks to support a wider variety of workloads, we have studied how current and emerging applications make use of these platforms. In this talk I will discuss some of our findings, including a description of our recent work on implementing a serverless architecture for digital mobility.

Bio: Cristina L. Abad is a Professor in the Department of Electrical Engineering and Computer Science at Escuela Superior Politécnica del Litoral in Guayaquil-Ecuador, where she leads the Distributed Systems Research Lab and co-directs the Big Data Research Group. She received her Ph.D. in 2014 from the University of Illinois at Urbana-Champaign. For three years during her PhD, she was a Software Engineering Intern in the Hadoop Core Team at Yahoo, where she worked on workload modeling and evaluation of the HDFS and had the opportunity to contribute to the Apache Hadoop codebase. Her research interests lie at the intersection of Distributed Systems and Performance Engineering. In particular, she works on designing and building distributed systems that can self-adapt to workload changes and maximize performance, with applications in cloud computing and Big Data. Her international funding sources have included VLIR-UOS, Google, Microsoft, Amazon Web Services, and AT&T Labs Research. She has received a Fulbright Fellowship, a UIUC CS Excellence Fellowship, and two Google Faculty Research Awards. Cristina is a member of IEEE, ACM, SPEC RG, and Usenix, and is the elected Secretary of the SPEC RG.