Banner Banner

November 09, 2020

Prof. Dr. Volker Markl

Major extension of the EDBT 2019 best paper by TU Berlin and DFKI researchers accepted for publication in TODS

Major Extension of the EDBT 2019 Best Paper by TU Berlin and DFKI Researchers Accepted for Publication in TODS

The paper “Scotty: General and Efficient Open-Source Window Aggregation for Stream Processing Systems” by J. Traub, P. M. Grulich, A. R. Cuéllar, S. Breß, A. Katsifodimos, T. Rabl, and V. Markl was accepted for publication at ACM Transactions on Database Systems (TODS).

This extended journal paper is a major extension of the EDBT best paper titled Efficient “Window Aggregation with General Stream Slicing” from 2019 by the same authors from the Database Systems and Information Management (DIMA) group at TU Berlin and the Intelligent Analytics for Massive Data (IAM) group at DFKI. Among other extensions, the new journal paper was extended with detailed algorithm specifications, API-examples, and examples for using Scotty in different streaming systems.


J. Traub, P. M. Grulich, A. R. Cuéllar, S. Breß, A. Katsifodimos, T. Rabl, and V. Markl

Window aggregation is a core operation in data stream processing. Existing aggregation techniques focus on reducing latency, eliminating redundant computations, or minimizing memory usage. However, each technique operates under different assumptions with respect to workload characteristics such as properties of aggregation functions (e.g., invertible, associative), window types (e.g., sliding, sessions), windowing measures (e.g., time- or count-based), and stream (dis)order. In this paper, we present Scotty, an efficient and general open-source operator for sliding-window aggregation in stream processing systems, such as Apache Flink, Apache Beam, Apache Samza, Apache Kafka, Apache Spark, and Apache Storm. One can easily extend Scotty with user-defined aggregation functions and window types. Scotty implements the concept of general stream slicing and derives workload characteristics from aggregation queries to improve performance without sacrificing its general applicability. We provide an in-depth view on the algorithms of the general stream slicing approach. Our experiments show that Scotty outperforms alternative solutions by up to one order of magnitude.