Spectrum crowdsensing is a paradigm where participants upload their collected spectrum data to the cloud for extracting analytics. First movers like Microsoft Spectrum Observatory and Electrosense, though with support from leading industry, research, and government, still suffer from sustainability challenges. In this paper, we present Fiesta, a sustainable framework for spectrum crowdsensing. On the technology side, we use federated learning and blockchain to decentralize the data analysis computations. For individual participants, minimal invasion of privacy suppresses concerns regarding large-scale adoption. From organizations’ perspectives, using blockchain avoids single point of failure and enhances the robustness of the entire system against malicious attacks. On the policy side, we propose a reward quantification mechanism to motivate engagement. Potential funding sources to ensure ongoing sustainability are also discussed. We have demonstrated Fiesta through simulation testbeds and real-world deployments with two demo tasks. Results show that Fiesta, as a decentralized framework, can preserve user privacy, enhance system robustness, maintain data fidelity compared with traditional methods, and fairly reward participants. We believe Fiesta is a stepping stone for the future spectrum crowdsensing paradigm.

Deploying large language models (LLMs) locally on mobile devices is advantageous in scenarios where transmitting data to remote cloud servers is either undesirable due to privacy concerns or impractical due to network connection. Recent advancements have facilitated the local deployment of LLMs. However, local deployment also presents challenges, particularly in balancing quality (generative performance), latency, and throughput within the hardware constraints of mobile devices. In this paper, we introduce our lightweight, all-in-one automated benchmarking framework that allows users to evaluate LLMs on mobile devices. We provide a comprehensive benchmark of various popular LLMs with different quantization configurations (both weights and activations) across multiple mobile platforms with varying hardware capabilities. Unlike traditional benchmarks that assess full-scale models on high-end GPU clusters, we focus on evaluating resource efficiency (memory and power consumption) and harmful output for compressed models on mobile devices. Our key observations include: i) differences in energy efficiency and throughput across mobile platforms; ii) the impact of quantization on memory usage, GPU execution time, and power consumption; and iii) accuracy and performance degradation of quantized models compared to their non-quantized counterparts; and iv) the frequency of hallucinations and toxic content generated by compressed LLMs on mobile devices.

RDMA datacenters are proliferating to meet the demand of emerging workloads such as AI training and inference as well as distributed storage. This trend has opened up a critical knowledge gap: the traffic characteristics of congestion in these networks remain unknown. We do not know, for example, which layers of the network are the most congested, if the network is load balanced effectively, how long congestion events last, and how accurate existing telemetry systems are in capturing congestion. This paper bridges this gap by investigating congestion in a large-scale RDMA datacenter dedicated to distributed AI training. We provide insights into three specific congestion patterns: (a) location and distribution in the network, (b) burstiness, e.g., the duration and synchrony of bursts, and (c) observability using existing telemetry methods. We show, for instance, that the deployment of Priority Flow Control (PFC) in RDMA networks has shifted the location of congestion one level up: from the edge-host in legacy TCP/IP datacenters to the network core in RDMA datacenters. At the same time, we show that the same protocol enables us to observe and understand congestion better, even bursty events. The findings of this research reveal open challenges for measuring, characterizing, and managing congestion in RDMA networks, paving the way for future research.

Radio frequency (RF) techniques have shown promise for continuous contactless healthcare applications. However, real-world indoor environments pose challenges for existing systems, which may struggle to detect subtle physiological signals. This paper proposes Medusa, a novel wireless vital-sign sensing system designed for multi-view setups. It enables users to deploy distributed Multiple Input Multiple Output (MIMO) arrays into their daily living environments, facilitating vital-sign sensing in real-world settings. Unlike most existing single Commercial Off-The-Shelf (COTS) radar-based systems that operate under controlled settings Medusa's primary novelty lies in the design of a first-of-its-kind flexible multi-view vital sign sensing system that is view-agnostic, pose-agnostic, contactless, and can sense basic human vitals with good accuracy. Through our well-engineered hardware and software co-design, Medusa enables real-time processing of large distributed MIMO arrays, while balancing the tradeoff between Signal-to-Noise Ratio (SNR) and spatial diversity gain across each of its four distributed 4 × 4 sub-arrays for increased robustness. This is achieved using our novel unsupervised learning model which effectively recovers vital sign waveforms by decomposing the received signals. Extensive evaluations with 21 participants demonstrate Medusa's spatial diversity gain for real-world vital-sign monitoring, enabling free movement and orientation of subjects in both familiar and unfamiliar indoor environments.

Understanding spectrum activity is challenging when attempted at scale. The wireless community has recently risen to this challenge in designing spectrum monitoring systems that utilize many low-cost spectrum sensors to gather large volumes of sampled data across space, time, and frequencies. These crowdsensing systems are limited by the uplink bandwidth available to backhaul the raw in-phase and quadrature (IQ) samples and power spectrum density (PSD) data needed to run various applications. This paper presents FlexSpec, a framework based on the Walsh-Hadamard transform to compress spectrum data collected from distributed and low-cost sensors for real-time applications. This transformation allows sensors to significantly save uplink bandwidth thanks to its inherent properties both when it is applied to IQ and PSD data. Additionally, by leveraging a feedback loop between the sensor and the edge device it connects to, FlexSpec carefully adapts the compression ratio over time to changes in the spectrum and different applications, jointly considering data size, application performance, and spectrum variations. We experimentally evaluate FlexSpec in several applications. Our results show that FlexSpec is particularly suitable for IoT transmissions and signals close to the noise floor. Compared with prior work, FlexSpec provides up to$7\times $more reduction of uplink data size for signal detection based on PSD data, and reduces up to$6\times $to$8\times $the number of undecodable messages for IQ sample decoding.

Multipath TCP (MPTCP) is a promising protocol that aggregates the bandwidth of mobile client's multiple interfaces. However, currently it is still not widely adopted. A key reason for this slow adoption is that the layer-4 load balancers (LBs) used to scale TCP based services in data centers are not MPTCP-aware and forward the multiple TCP subflows of the same MPTCP connection independently to different backends (BEs). In this paper, we present RomanRoads (RR), an MPTCP-aware layer-4 LB which eliminates this hurdle to widespread MPTCP adoption. Compared with prior proposals, RR is easily deployable because it does not change the service provider's network configuration, makes no modification to ordinary TCP protocol and only minimal modification to the MPTCP connection setup process, and supports the case of multiple LBs. We implement RR in the form of a software LB and validate its correctness and high performance through extensive experiments. RR achieves 100% correctness at steady state, no connection disruption during LB churns, and line-rate throughput for packets from 5-tuple seen before. Moreover, we shed light on the desired properties of an LB-friendly multipath layer-4 protocol to provide guidance for future multipath protocol design.

Understanding spectrum characteristics with little prior knowledge requires fine-grained spectrum data in the frequency, spatial, and temporal domains; gathering such a diverse set of measurements results in a large data volume. Analysis of the resulting dataset poses unique challenges; methods in the status quo are tailored for specific spectrum-related applications (apps), and are ill equipped to process data of this magnitude. In this paper, we design BigSpec, a general-purpose framework that allows for fast processing of apps. The key idea is to reduce computation costs by performing computation extensively on compressed data that preserves signal features. Adhering to this guideline, we build solutions for three apps, i.e., energy detection, spatio-temporal spectrum estimation, and anomaly detection. These apps were chosen to highlight BigSpec's efficiency, scalability, and extensibility. To evaluate BigSpec's performance, we collect more than 1 terabyte of spectrum data spanning a year, across 300MHz-4GHz, covering 400 km2. Compared with baselines and prior works, we achieve 17× run time efficiency, sublinear rather than linear run time scalability, and extend the definition of anomaly to different domains (frequency & spatio-temporal). We also obtain high-level insights from the data to provide valuable advice on future spectrum measurement and data analysis.

Data compression is a popular technique for improving the efficiency of data processing workloads such as SQL queries and more recently, machine learning (ML) with classical batch gradient methods. But the efficacy of such ideas for mini-batch stochastic gradient descent (MGD), arguably the workhorse algorithm of modern ML, is an open question. MGD's unique data access pattern renders prior art, including those designed for batch gradient methods, less effective. We fill this crucial research gap by proposing a new lossless compression scheme we call tuple-oriented compression (TOC) that is inspired by an unlikely source, the string/ text compression scheme Lempel-Ziv-Welch, but tailored to MGD in a way that preserves tuple boundaries within mini-batches. We then present a suite of novel compressed matrix operation execution techniques tailored to the TOC compression scheme that operate directly over the compressed data representation and avoid decompression overheads. An extensive empirical evaluation with real-world datasets shows that TOC consistently achieves substantial compression ratios by up to 51x and reduces runtimes for MGD workloads by up to 10.2x in popular ML systems.

This paper presents the first comprehensive model of the 802.1laa intra-access category prioritization under saturated condition. We propose a 3D Markov chain to model the system and analyze several important QoS metrics, including throughput, delay, and frame drop ratio. The accuracy of our model is verified by extensive simulations. The results show that the 802.1laa can provide a finer grained prioritization between individual audio and video streams in throughput and delay. However, advanced physical layer techniques are necessary in providing QoS support in a large-scale WiFi network.