A new research paper by Red Hat associate Ilya Kolchinsky has recently been accepted to ACM SIGMOD ‘22, a leading international conference in the field of large-scale data management systems and databases. Running annually since 1975 with a highly selective paper acceptance rate of 17%, SIGMOD is a top-tier forum for database researchers and developers to explore the latest state-of-the-art advances and results in the area.
The paper, titled “DLACEP: A Deep-Learning Based Framework for Approximate Complex Event Processing”, is co-authored by Prof. Assaf Schuster, one of the leading academic partners of Red Hat Research, as well as Adar Amir, a graduate student supervised by Assaf and Ilya. It presents a major breakthrough in the area of efficient real-time processing of massive live data streams. For the paper, Adar, Ilya, and Assaf studied the ways to incorporate modern deep-learning methodologies into the online stream pattern detection process (also known as complex event processing, or CEP) to achieve performance boosts of multiple orders of magnitude. By utilizing the approaches proposed in the paper, systems performing computation-heavy on-the-fly stream processing and analysis could dramatically improve throughput and latency while at the same time reducing peak memory consumption. As the first work to consider the fusion of deep learning and CEP, this paper paved the way for future research in this highly innovative field.
The paper will be presented at the SIGMOD conference, which will be held June 12-17 in Philadelphia, PA, USA, and will appear in the official proceedings.The novel ideas discussed in the paper could be implemented in a future version of OpenCEP— an advanced open source complex event processing framework with cutting-edge pattern detection capabilities. This project, jointly led by Ilya and Assaf, is in the concluding stages of development and will be available for use in the coming months.
