Modular real-time streaming architecture for high-throughput DAS monitoring

Distributed Acoustic Sensing (DAS) converts fiber-optic cables into continuous seismic sensors, producing measurements at rates exceeding hundreds of megabytes per second. Conventional workflows designed for periodic analysis introduce a latency incompatible with real-time monitoring: events remain invisible until the next scheduled processing window completes, often hours or days after occurrence. Furthermore, this process is brittle for monitoring contexts as data is processed with predefined parameters, and changing those requires manual effort. Additionally, reprocessing historic data of those volumes requires a lot of time and compute, either of which may not be readily available when the need occurs. To address those challenges, we present a modular, real-time system architecture consisting of interconnected services designed for high throughput data ingestion and processing. The system will be deployed within the POSEIMON‑I project, which focuses on extending induced seismicity monitoring in the vicinity of the Porthos CO₂ storage site. It is built around two core components - Apache Kafka for data streaming and Apache Flink for stateful, real-time processing of live data from multiple sources. Additionally, further services allow for the storage, querying and visualization of operational, processed or derived event data. Additionally, algorithms can be deployed or removed from a running system through task-based orchestration enabling the continuous development, refinement and evaluation of algorithms that can adapt to a changing environment. This system design bridges the gap between research and industry by enabling the use of industry-grade systems in scientific contexts and bringing experimentation closer to real-world deployments. Finally, the system is fully built on open-source components ensuring data and system sovereignty and albeit designed to deal with high-throughput data, characteristic for DAS, is agnostic to the domain and can be applied in any context with live data-streams.