We are happy to announce that three papers got accepted at EDBT 2025! The papers will be presented as part of the EDBT conference in Barcelona, Spain on 25th of March - 28th of March, 2025.
1) An Empirical Evaluation of Serverless Cloud Infrastructure for Large-Scale Data Processing by Thomas Bodner, Theo Radig, David Justen, Daniel Ritter, and Tilmann Rabl
Abstract:
Data processing systems are increasingly deployed in the cloud. While monolithic systems run fully on virtual servers, recent systems embrace cloud infrastructure and utilize the disaggregation of compute and storage to scale them independently. The introduction of serverless compute services, such as AWS Lambda, enables finer-grained and elastic scalability within these systems. Prior work shows the viability of serverless infrastructure for scalable data processing, but sees limitations due to performance variance and cost overhead, especially in networking and storage. In this paper, we perform a detailed analysis of the performance and cost characteristics of serverless infrastructure in the data processing context. We base our analysis on a large series of microbenchmarks across different compute and storage services, as well as end-to-end workloads. To enable our analysis, we propose the Skyrise serverless evaluation platform. For the widely used serverless infrastructure of AWS, our analysis reveals distinct boundaries for performance variability in serverless networks and storage. We also present cost break-even points for serverless compute and storage. These insights provide guidance on when and how serverless infrastructure can be used efficiently for data processing.
2) An Interactive Analysis of Serverless Cloud Infrastructureby Thomas Bodner and Tilmann Rabl
Abstract:
Data processing systems are increasingly being deployed in the cloud, because of the benefits of elasiticity and short-term resource provisioning. In recent years, serverless cloud computing is offered in the form of highly elastic resource pools. However, analyzing and understanding the performance and cost characteristics of serverless cloud infrastructure in the context of data processing is challenging. In this demonstration, we present our Skyrise evaluation framework for experimentation in serverless data processing. The framework provides a suite of micro-benchmarks and a serverless query engine to run end-to-end workloads. Users can benchmark serverless compute and storage resources and observe the impact of their performance on data-intensive applications. In addition, they can study the trade-offs compared to server-based systems. Utilizing the inherent elasticity of serverless resources, our framework facilitates interactive analysis and supports the understanding of key characteristics of serverless data processing for data system engineers.
3) Dema: Efficient Decentralized Aggregation for Non-Decomposable Quantile Functions by Wang Yue, Martin Boissier, Manisha Luthra, and Tilmann Rabl
Abstract:
The growing number of Internet of Things (IoT) devices has led to the widespread adoption of decentralized networks to handle unbounded data streams in a variety of applications. Traditional stream processing engines rely on centralized window aggregation, resulting in high network overhead and processing bottlenecks. Current decentralized solutions mitigate these issues by offloading partial aggregations to edge devices, but they only support decomposable functions like sum and count. Non-decomposable functions, such as median and quantile, remain a challenge as partial results cannot be merged without accessing the complete dataset. To address this, we propose Dema, a decentralized window aggregation technique for non-decomposable functions. Dema reduces network traffic and computational load by performing localized sorting and transmitting statistical summaries rather than raw data. Our approach efficiently calculates median and quantile values, achieving up to a 99% reduction in network traffic compared to state-of-the-art methods. Our evaluation results show that Dema significantly outperforms existing approaches in terms of throughput and scalability, while ensuring accurate results.
