SearcharxivSearch

arXiv subjects

Berk Buzcu

Publications and source records attributed to Berk Buzcu.

2 recordsLinked to original sources

Modular Multi-Domain Digital Twin Architecture: Sustainable Intent-Driven 6G Management

Future 6G networks will operate across distributed and heterogeneous domain infrastructures, making conventional single-domain management insufficient for proactive, trustworthy automation. Network Digital Twins (NDTs) enable what-if analysis, AI-assisted optimization, and risk-free validation of control actions before deployment, yet monolithic end-to-end twins remain impractical due to scalability, fidelity, and cross-domain coordination challenges. Accordingly, this paper proposes a Digital Twin-enabled 6G architecture that exposes NDT capabilities as a specialized service domain within a multi-domain orchestration framework built on a state-of-the-art service-based 6G architecture. A DT Orchestrator interprets \textit{predictive} and \textit{prescriptive} what-if queries and composes domain-specific DT modules and simulators on demand, while decision authority remains with the requesting entity. Furthermore, a generalized workflow covers telemetry synchronization, simulation-based decision support, and closed-loop execution. The framework is demonstrated through a green-networking use case that couples a system-level O-RAN cellular digital twin component with a two-stage solar-allocation simulator, evaluated over a 105-base-station deployment in Poznan using simulative datasets. Joint coverage and renewable optimization reduces daily grid consumption by 28.5\% with 32 solar panels at the diminishing-returns threshold, with 17 base stations identified as both coverage-active and high-priority solar candidates as evidence that cross-domain NDT coordination enables sustainable, intent-driven 6G network management.

cs.NI

Sailor: Automating Distributed Training over Dynamic, Heterogeneous, and Geo-distributed Clusters

The high GPU demand of ML training makes it hard to allocate large homogeneous clusters of high-end GPUs in a single availability zone. Leveraging heterogeneous GPUs available within and across zones can improve throughput at a reasonable cost. However, training ML models on heterogeneous resources introduces significant challenges, such as stragglers and a large search space of possible job configurations. Current systems lack support for efficiently training models on heterogeneous resources. We present Sailor, a system that automates distributed training over heterogeneous, geo-distributed, and dynamically available resources. Sailor combines an efficient search space exploration algorithm, accurate runtime and memory footprint simulation, and a distributed training framework that supports different types of heterogeneity to optimize training throughput and cost.

cs.DC