Tutor: Seiwerth
Distributed co-simulation frameworks rely on message-based communication for data exchange between simulation participants, participants synchronization, and simulation orchestration. The underlying messaging platform for these frameworks must provide efficient communication and accumulate less platform overhead to achieve high performance in co-simulations execution.
DaceDSX is currently deployed using Confluent platform with Apache Kafka providing topic based publish/subscribe messaging. The study identifies the core services required for executing DaceDSX co-simulations and evaluates multiple Apache Kafka compatible messaging architectures using deterministic benchmark scenarios. A comparative analysis is presented in terms of simulation run time, message overhead, resources consumption, and scalability. The results show that existing Confluent platform can be replaced by minimal Kafka deployment configuration without affecting DaceDSX simulations. KRaft mode Kafka was also found to be a suitable alternative, whereas Redpanda performed competitively for smaller workloads but could not be scaled for larger or parallel workloads.
Room 04.137, Martensstr. 3, Erlangen
or
Zoom:
https://fau.zoom-x.de/j/68350702053?pwd=UkF3aXY0QUdjeSsyR0tyRWtLQ0hYUT09
Meeting-ID: 683 5070 2053
Code: 647333