AI & ML
OrbitWeaver: LLM Agent Planning with Simulation Feedback for Optically Interconnected Orbital GPU Clusters
arXiv eess.SY0 views
arXiv:2611.17850v1 Announce Type: cross Abstract: Space-based data-center designs increasingly consider GPU computing formations connected by inter-satellite optical links. With limited neighboring connections, jobs assigned disjoint GPUs can still contend on shared relay paths. Rank placement and communication priorities jointly affect this contention and the readiness of downstream computation. We present OrbitWeaver, an offline large language model (LLM) agent planner that jointly optimizes rank placement and communication order. Guided by communication structure, execution dependencies, and simulation feedback, the agent constructs and refines plans that coordinate shared-link access with computation progress. This coordination reduces contention and dependency blocking, shortening mean job completion time (JCT). Simulation experiments on optically interconnected orbital GPU clusters show that OrbitWeaver reduces aggregate mean JCT by 6.2-33.5% relative to the state-of-the-art baselines CASSINI, Crux, and Symphony. These results demonstrate the value of LLM-guided planning for efficient concurrent computing in orbital GPU clusters.
Source:https://arxiv.org/abs/2611.17850PreviousMethanol substitution suppresses deflagration-excited knock and increases the efficiency of a hydrogen-fuelled CFR engineNextSafety Certificates for Systems under Attack via Stackelberg Set Games
Related posts
Goodbye, Retro-Futurism
0dev.to
Turn One Broad AI Topic Into Five Useful Reader Questions
0dev.to
Raw public keys move TLS identity into provisioning and rotation
0dev.to Security
External PSK importers make TLS key reuse a deployment decision
0dev.to Security

Why VirtIO Is Faster Than Emulated Devices: Virtqueues and vhost-net Explained
0dev.to Linux
HASHCAT
0dev.to Linux