Searcharxiv⌕ Search

arXiv subjects

Roger Sanchez-Vital

Publications and source records attributed to Roger Sanchez-Vital.

5 recordsLinked to original sources

Flux: Optimal Scheduling of Optical Circuit Switches for LLM Training

Optical Circuit Switching (OCS) offers high bandwidth density and energy efficiency for LLM training, but incurs a non-negligible reconfiguration delay. Prior work typically schedules optical circuit switches independently of compute, using aggregate traffic demand to determine which circuits to provision and when. We argue that this separation creates a fundamental inefficiency: reconfigurations that ignore the compute timeline can stall communication, resulting in low circuit utilization and large buffer requirements. In this paper, we present Flux, a scheduler that optimally schedules optical circuit switches based on the structure of the entire workload. Flux remains effective across a wide range of switching speeds by reusing circuits and amortizing reconfiguration delay behind compute and communication. We show that Flux reduces training iteration time by up to $10\times$ and peak NIC buffer requirements by more than three orders of magnitude compared to traditional periodic schedulers.

cs.NI↗

Disaggregated LLM Inference in Space: A Communication Feasibility Study

Orbital data centers could enable large-scale AI inference, but efficiently serving a frontier language model to many users may require several compute satellites to cooperate over optical links. While disaggregated inference, which splits large language model (LLM) serving across machines, is well established in terrestrial data centers, its viability under the unique constraints of orbital deployments remains an open question. This article investigates how link capacity and satellite separation affect two types of disaggregated LLM serving, in both compact satellite formations and distributed constellations. Prefill-decode disaggregation (PDD) migrates a large KV cache once per request, while attention-FFN disaggregation (AFD) exchanges small activations at every mixture-of-experts layer for every generated token. The two respond to distance in different ways. PDD is limited mainly by link capacity and, with fast enough links, tolerates separations of up to a thousand kilometers. AFD is limited mainly by propagation delay, which accumulates layer by layer and forces attention and expert instances within about a kilometer of each other. Based on these results, we propose a hierarchical orbital inference architecture and provide network design guidelines for future orbital data centers, including the role of ground access latency.

cs.NI↗

LEO Constellation Analysis for Satellite IoT: An Analysis of Coverage Gaps and Retransmissions

Low Earth Orbit (LEO) constellations are a promising way to connect Internet of Things (IoT) devices beyond terrestrial coverage, yet the intermittent visibility of their satellites imposes hard constraints on system and protocol design. Coverage probability and other snapshot statistics, which are often computed without accounting for the terminal's latitude, are insufficient for this purpose. Designers also need the temporal structure of coverage: the rate of coverage gaps, their duration distribution and mean, and quantities such as the probability that a retransmission succeeds after an initial transmission has failed. We develop an ergodic framework for Walker Star constellations in which these quantities reduce to spatial integrals over a two-dimensional torus that can be evaluated efficiently. We validate the framework against a seven-day Simplified General Perturbations 4 (SGP4) simulation of the Iridium NEXT constellation and find close agreement for the coverage probability, the revisit-time distribution, and the retransmission success probability. A comparison of Walker Star configurations further shows that constellations with similar satellite counts can yield substantially different packet-delivery probabilities with at most one retransmission, depending on the number of orbital planes. The results show that the ergodic model accurately gives the long-run values that finite time-domain simulations can only approximate, providing researchers with a tractable basis for the analytical modelling of IoT connectivity over non-terrestrial networks (NTN).

cs.IT↗

A Primer on AP Power Save in Wi-Fi 8: Overview, Analysis, and Open Challenges

Wi-Fi facilitates the Internet connectivity of billions of devices worldwide, making it an indispensable technology for modern life. Wi-Fi networks are becoming significantly denser, making energy consumption and its effects on operational costs and environmental sustainability crucial considerations. Wi-Fi has already introduced several mechanisms to enhance the energy efficiency of non-Access Point (non-AP) stations (STAs). However, the reduction of energy consumption of APs has never been a priority. Always-on APs operating at their highest capabilities consume significant power, which affects the energy costs of the infrastructure owner, aggravates the environmental impact, and decreases the lifetime of battery-powered APs. IEEE 802.11bn, which will be the basis of Wi-Fi 8, makes a big leap forward by introducing the AP Power Save (PS) framework. In this article, we describe and analyze the main proposals discussed in the IEEE 802.11bn Task Group (TGbn), such as Scheduled Power Save, (Semi-)Dynamic Power Save, and Cross-Link Power Save. We also consider other proposals that are being discussed in TGbn, namely the integration of Wake-up Radios (WuRs) and STA offloading. We then showcase the potential benefits of AP PS in several scenarios, including a deployment of 470 real APs in a university campus. Our numerical analysis reveals that AP power consumption can be decreased on average by up to 28 percent, with further improvement potential. Finally, we outline the open challenges that need to be addressed to optimally integrate AP PS in Wi-Fi and ensure its compatibility with legacy devices.

cs.NI↗

Energy performance of LR-FHSS: analysis and evaluation

Long-range frequency hopping spread spectrum (LR-FHSS) is a pivotal advancement in the LoRaWAN protocol that is designed to enhance the network's capacity and robustness, particularly in densely populated environments. Although energy consumption is paramount in LoRaWAN-based end devices, {this is the first study} in the literature, to our knowledge, that models the impact of this novel mechanism on energy consumption. In this article, we provide a comprehensive energy consumption analytical model of LR-FHSS, focusing on three critical metrics: average current consumption, battery lifetime, and energy efficiency of data transmission. The model is based on measurements performed on real hardware in a fully operational LR-FHSS network. While in our evaluation, LR-FHSS can show worse consumption figures than LoRa, we find that with optimal configuration, the battery lifetime of LR-FHSS end devices can reach 2.5 years for a 50 min notification period. For the most energy-efficient payload size, this lifespan can be extended to a theoretical maximum of up to 16 years with a one-day notification interval using a cell-coin battery.

cs.NI↗