Technical updates, experimental findings, system design notes and research insights from the DDLSim-Lab project.
Explore recent technical work covering distributed-system failures, intelligent scheduling and realistic edge AI simulation.
DDLSim-Lab's fault injection capabilities have been expanded to support more sophisticated distributed-system failure models, including Byzantine behavior, network partitions and cascading failures.
In our latest research, we've enhanced DDLSim-Lab's fault injection capabilities to support more sophisticated failure models. Beyond traditional latency and packet loss, we now simulate Byzantine failures, network partitions and cascading failures. These advanced fault models enable researchers to test distributed training algorithms under adverse conditions.
We've implemented a reinforcement learning framework within DDLSim-Lab that dynamically allocates computational and network resources across heterogeneous edge-cloud infrastructures. The system treats each node as an agent in a multi-agent reinforcement learning environment.
DDLSim-Lab now supports realistic simulation of edge AI scenarios where constrained devices with intermittent connectivity participate in federated learning processes. Researchers can define custom device profiles representing real-world IoT deployments.
Try another search term or research category.
Failure models, resilience, recovery behavior and distributed system reliability.
Intelligent resource allocation, reinforcement learning and heterogeneous workloads.
Latency, packet loss, network behavior and communication bottlenecks.
Resource-constrained devices, edge-cloud environments and distributed inference.
Device heterogeneity, aggregation, stragglers and adaptive training.
Convergence, accuracy, recovery time and system-level measurements.
Explore the research direction, technical documentation and source code behind DDLSim-Lab.