As demand grows, serverless computing systems must scale to meet increasing throughput requirements. Cloud computing easily achieves scalability by allocating abundant and elastic resources. In contrast, edge computing pre-deploys scarce and inelastic resources on site. However, edge applications often need to scale dramatically to handle bursty demand. Because they typically serve fewer users with highly variable workloads, our study shows that peak usage may require up to 8× more resources to be pre-deployed across edge nodes than in a centralized cloud. We observe that a typical serverless request can often be satisfied by different implementations, with dissimilar resource consumption profiles. When an edge application fails to scale up, the culprit is often a single bottleneck resource being fully consumed, with other resources readily available. Motivated by this observation, we introduce SCALEWAVE, a middleware for seamlessly scaling up with different implementations to fully utilize all available resources to achieve scalable serverless computing at the edge. Supporting multiple implementations, however, introduces new challenges for conventional auto-scaler designs. Reactive strategies tend to yield suboptimal performance, while proactive methods struggle with compatibility. SCALEWAVE overcomes these limitations by proactively distributing traffic across implementations, while leveraging existing autoscalers to manage instance scaling reactively for each one. Our evaluations using real workload traces on a heterogeneous cluster of edge devices demonstrate that our design allows edge applications to serve 2x more requests with minimal latency and achieve 50% more successful requests during workload bursts — showcasing substantial gains in scalability and resource efficiency.

Web services are widely used in modern software, providing diverse data and functionalities. Some data and functionalities are critical to an application's execution and user experience, posing strict requirements on the Quality of Service (QoS) of their delivery (e.g., latency and reliability), which services often fail to meet. Previous studies show that composing homogeneous services, i.e., simultaneously invoking multiple services providing the same functionalities and returning the first response, can improve latency and reliability. However, this approach increases the workloads on cloud servers and causes additional network traffic, limiting its deployment at scale. Our empirical study reveals that services deliver varying QoS across different locations, making it possible to reduce the invocation cost by tailoring the composition strategy for different clients. In this paper, we introduce an approach that composes homogeneous services dynamically for each client, improving user-perceived QoS while minimizing the invocation costs. In particular, our approach first probes the QoS of all homogeneous services for a client, and then calculates an optimal composition strategy that satisfies the QoS requirements specified by App developers with minimum cost. We prototyped our approach as an Android library and tested it via both real-world experiments and simulations. The evaluation results show that our approach significantly improves QoS compared to invoking a single service with average best QoS across all locations (enhancing reliability to 100%, reducing average latency by 7% and tail latency by 35%) while incurring 50% less cost than static homogeneous composition, making it a useful tool for service-oriented applications.

Recent advancements in Large Language Models and Generative Artificial Intelligence have significantly boosted the usage of Machine Learning (ML) services, leading to a variety of providers offering comparable services that have the same functionalities but differ in their model structures and training datasets. Application developers face the challenge of selecting the most accurate ML service for their applications, which necessitates effectively measuring the accuracy of these equivalent services using application-specific data as inputs. This process is both time-consuming and labor-intensive, as it requires developers to manually label the expected outcomes for each input. In this paper, we formalize and address the important but long-overlooked “sample selection problem;’ which involves selecting and labeling a subset of inputs to measure the accu-racy of services with minimal error. Our Collaborative Online Sample Selection (COSS) approach considers the correlation of the invocation results from all candidate services for all inputs when selecting samples. Specifically, COSS clusters inputs into “patterns” based on their service invocation outcomes and employs an algorithm to iteratively select inputs based on each pattern's contribution to the overall accuracy measurement, and the confidence in the impact of the already-selected inputs on the accuracy measurement. We measure the effectiveness of our algorithm on three sets of web services, and our results indicate that our algorithm significantly reduces measurement errors by an average of 54% and labeling effort by an average of 30%.

Serverless computing abstracts away server management, enabling automatic scaling, efficient resource utilization, and cost-effective pricing models. However, despite these advantages, it faces the significant challenge of cold-start latency, adversely impacting end-to-end performance. Our study shows that many serverless functions initialize libraries that are rarely or never used under typical workloads, thus introducing unnecessary overhead. Although existing static analysis techniques can identify unreachable libraries, they fail to address workload-dependent inefficiencies, resulting in limited performance improvements. To overcome these limitations, we present SlimStart, a profile-guided optimization tool designed to identify and mitigate inefficient library usage patterns in serverless applications. By leveraging statistical sampling and call-path profiling, SlimStart collects runtime library usage data, generates detailed optimization reports, and applies automated code transformations to reduce cold-start overhead. Furthermore, SlimStart integrates seamlessly into CI/CD pipelines, enabling adaptive monitoring and continuous optimizations tailored to evolving workloads. Through extensive evaluation across three benchmark suites and four real-world serverless applications, SlimStart achieves up to a 2.30× speedup in initialization latency, a 2.26× improvement in end-to-end latency, and a 1.51× reduction in memory usage, demonstrating its effectiveness in addressing cold-start inefficiencies and optimizing resource utilization.

As concerns around end-user privacy continues to grow, Federated Recommendation Systems (FRS) have emerged as a privacy-preserving alternative to traditional cloud-based recommendation systems. They keep personal data "always mine"—never leaving the user's device during both training and inference of the recommendation models. However, our empirical study shows that phone-based FRS often suffer from poor recommendation accuracy. This is because, in real-world settings, only a small fraction of phones can contribute data to model updates, as they are often not charging or connected to Wi-Fi when training occurs. In this paper, we propose using "always-on" personal home routers—devices that are continuously powered and consistently network-connected—as a promising alternative to intermittently available mobile phones for FRS. Despite its potential, router-based FRS introduces two key challenges. First, routers have more limited and variable computational resources than phones, leading to slower model convergence during federated training, as the system often waits on slower but data-rich devices. Second, routers introduce an additional layer of distributed infrastructure, increasing the complexity of managing training and inference workflows for application developers. To address these challenges, we present a middleware framework for router-based FRS that includes: (1) a customized federated training strategy that maximizes data utilization from slower routers without significantly increasing training time, and (2) a declarative programming model that simplifies the integration of router-based FRS into applications. Our evaluation shows that the router-based FRS improves recommendation accuracy by up to 53% compared to phone-based systems, while our framework reduces convergence time by up to 43% relative to state-of-the-art federated training methods, with minimal programming effort and runtime overhead.

Fine-grained bird sighting prediction is crucial for advancing ecological research, informing conservation planning, and enhancing the birdwatching experience while fostering public awareness of biodiversity. The rapid expansion of citizen-based bird observation networks has led to an exponential accumulation of bird sighting records, which can be leveraged to train machine learning models for more precise predictions. However, general-purpose machine learning models often fail to incorporate ecological factors that influence bird activity, resulting in less accurate predictions. In this paper, we present an ecologically informed machine learning framework based on LightGBM that integrates spatiotemporal correlations, ecological context, and dynamic environmental variables to improve bird sighting predictions. The framework captures temporal trends using rolling windows, applies spatial smoothing to account for observation proximity, and models ecological dependencies—such as temperature-food interactions—through interaction terms. Key environmental factors, including habitat classifications, weather conditions, and seasonally adjusted food availability proxies, are dynamically incorporated to enhance ecological relevance. Evaluation results demonstrate significant improvements in predictive accuracy, with increased F1-scores compared to baseline methods. By embedding ecological principles into machine learning models, this framework enables data-driven insights that reflect real-world environmental complexities, providing a powerful tool for biodiversity monitoring and conservation strategies.

While driving, drivers often multitask by interacting with the In-Vehicle Infotainment System (IVIS) to manage various applications, including map navigation and music playback. These tasks, requiring precise control, can distract drivers and increase the risk of road accidents. In this paper, we intro-duce WatchNavi, a system that enables drivers to control IVIS quantitatively with minimal distraction. WatchNavi leverages the increasing prevalence of smartwatches and utilizes the motion sensors in these devices to quantize control inputs, such as zoom levels and navigation commands (e.g., moving right or left). To enhance accuracy across different drivers, WatchNavi collects small amount of driver-specific motion data for specific control functions. This data is then used to train a lightweight Convolutional neural network (CNN) model for gesture recognition. We have implemented WatchNavi and assessed its performance in a real-world driving testbed. Our findings show that WatchNavi, powered by our refined gesture recognition model, significantly reduces driver distraction by 44.2% and improves the overall driving experience and safety.

Abstract Despite the widespread adoption of Web services in modern computing applications, there remains a lack of a systematic approach that can guide service developers in creating appealing services. This article addresses this gap by presenting findings from a comprehensive study of RapidAPI web services, the largest service marketplace, and their integration into GitHub-hosted applications. We collected data on over 16K RapidAPI services and 19K corresponding GitHub repositories invoking these services, evaluating each service based on metrics such as latency, reliability, pricing, followers, aggregate ratings community support, and provider support. Our analysis examines how these metrics influence service popularity and usage patterns on GitHub. We manually analyzed 800 GitHub repositories and identified developers’ service selection preferences and integration patterns, considering alternative services and their features. We then classified GitHub developers based on proficiency levels to understand how developers’ levels of proficiency impact their service selection and integration strategies. Moreover, we examined the metrics influence for matured set of repositories by excluding those intended solely for practice purposes. Our findings offer insights for service marketplaces to recommend integration-friendly services and for service developers to create offerings tailored to real-world application needs.

Service registry, a key component of the service-oriented architecture (SOA), aids software developers in discovering services that meet specific functionality requirements. Recent years have witnessed the transition from the traditional service registries to its successor, the Service Marketplaces, which involves deeper engagement in the SOA software lifecycle and offers additional features, such as service request delegation and monitoring of services' Quality of Service (QoS). However, by analyzing developers' questions posted on online Q&A forums, we found that many developers struggle with such transition, leading to development inefficiencies and even security vulnera-bilities. This paper presents the first empirical study aimed at uncovering the issues developers face with marketplaces, particularly those arising from the transition. Through a meticulous process of manually labeling and analyzing developers' questions, we develop a taxonomy of these issues, summarize the impacts caused by the transition, and provide actionable suggestions to App developers, service providers, and marketplaces. Utilizing the labeled questions and our insights, we fine-tune a Large Language Model (LLM) for providing answers to similar questions raised by developers and helping service providers and marketplaces extract useful information from these questions, such as service outages and key leakages. Our evaluation of the model's performance in answering and extracting pertinent information from a set of real-world questions demonstrates its effectiveness: it accurately classified 85 % of the queries and successfully identified 88 % of service names and 77 % of key leakages. As the first empirical study in this domain, this work not only aids developers in navigating the transition more effectively but also sheds light on the under explored issue of service registry evolution, offering valuable insights for researchers.

With the emergence of the open API ecosystem, third-party developers can publish their APIs on the API marketplace, significantly facilitating the development of cutting-edge features and services. The RapidAPI platform is currently the largest API marketplace and it provides over 40,000 APIs, which have been used by more than 4 million developers. However, such open API also raises security and privacy concerns associated with APIs hosted on the platform. In this work, we perform the first large-scale analysis of 32,089 APIs on the RapidAPI platform. By searching in the GitHub code and Android apps, we find that 3,533 RapidAPI keys, which are important and used in API request authorization, have been leaked in the wild. These keys can be exploited to launch various attacks, such as Resource Exhaustion Running, Theft of Service, Data Manipulation, and User Data Breach attacks. We also explore risks in API metadata that can be abused by adversaries. Due to the lack of a strict certification system, adversaries can manipulate the API metadata to perform typosquatting attacks on API URLs, impersonate other developers or renowned companies, and publish spamming APIs on the platform. Lastly, we analyze the privacy non-compliance of APIs and applications, e.g., Android apps, that call these APIs with data collection. We find that 1,709 APIs collect sensitive data and 94% of them dont provide a complete privacy policy. For the Android apps that call these APIs, 50% of them in our study have privacy non-compliance issues.

This research-to-practice paper introduces a mini-course module designed to teach computer science students how to interact more efficiently with Generative AI(GAI). The rapid rise of GAI is transforming education by providing students with easy access to knowledge and answers to their questions, acting as a personal tutor. Particularly in the field of computer science, where GAI can easily generate code based on specific requirements, many instructors struggle to prevent students from using tools like ChatGPT for completing assigned programming assignments and homeworks. However, we argue that 1) the use of GAI is inevitable, necessitating a redesign of courses so that students cannot merely rely on GAI without actual learning; and 2) students' learning can be enhanced if they learn to use GAI more effectively. In this paper, we demonstrate how we integrate Project-Based Learning to design the course module in a concise yet effective manner, which not only facilitates students' learning of GAI but also enriches their learning in relation to the host course where this mini-course module is embedded. In particular, the goal of this module is to teach CS students: 1) the basic principles and workflow of GAI; 2) Prompt Engineering: how to craft questions to interact more effectively with GAI; and 3) Extending GAI: how to create interactive tools by training customized GAI models. Designed to be completed within two weeks, the mini-course module can easily be incorporated into host courses. This mini-course module was integrated into a graduate-level Artificial Intelligence course with 42 students in Winter 2024. To assess the module's impact on student learning and engagement, we conducted pre- and post-course surveys as well as student interviews. The results from the surveys and interviews highlighted key areas for improving the design of educational modules to better teach essential GAI skills. These insights focused on enhancing student engagement and learning efficiency within a concise time frame.

Modern applications often invoke web services to access remote data and functionalities. The current service-oriented paradigm is “one-size-fits-all,” where App developers expect a single service to deliver satisfying Quality of Service (QoS) to all geographically and temporally dispersed clients. However, our empirical study reveals that despite the pervasive use of CDN and edge computing, many web services deliver significantly varied QoS to different users, resulting in some clients suffering from poor user experience. This paper introduces service polymorphism, a novel software paradigm that serves dispersed clients dissimilarly to improve their perceived QoS. Service polymorphism allows a client to maintain a list of equivalent services and invokes the one that offers the optimal QoS in the invocation context. The main challenge in supporting service polymorphism lies in minimizing the overhead for fine-grained QoS sensing. To address this challenge, we propose an edge- based QoS sharing mechanism that aggregates the context-specific QoS in edge servers, and allows clients to retrieve the QoS from local WiFi Access Points with minimized latency to decide the optimal service. Our evaluation shows that service polymorphism improves QoS significantly for 8 services out of 20, reducing their average latency by 231 ms (45%), tail latency by 80 ms (12 %), and error ratio from 0.2 % to 0 %.

To achieve low execution latency, time-sensitive applications, including AR/VR and autonomous driving, cache data at the edge of the network, close to end users. However, existing edge caches often fail to deliver low latency due to the inefficiency of DNS requests and the physical remoteness of their users. The solution described herein addresses these inefficiencies by presenting a millisecond-level, lightweight caching architecture that operates directly on widely deployed WiFi access points (APs). Specifically, our architecture interposes another level of caching closer to the client and is fine-tuned for APs's limited cache memory. Our solution (1) features a novel algorithm for managing cache at the AP level; (2) allows the cache query workflow to proceed at full speed; and (3) requires no changes to the application logic. Our evaluation demonstrates that our reference implementation can decrease application-level latency by as much as 76% compared to the existing solutions, without impacting AP core functions. Our caching architecture effectively improves application responsiveness by tapping into existing networking infrastructure, thus offering a powerful and cost-efficient system component for building emerging time-sensitive applications at the edge.

Surveillance cameras, even with Pan, Tilt, and Zoom (PTZ) capabilities, can only cover a limited directional range at a time, leading to inevitable blind spots. To mitigate these blind spots, users typically need to deploy additional cameras or sensors (such as motion sensors and microphones) to guide the PTZ platform’s movement for capturing intruders, incurring additional deployment and maintenance costs. To address this challenge, this paper introduces a novel approach that utilizes the camera’s built-in WiFi module to detect potential intruders and direct the PTZ platform’s movements. The approach involves collecting WiFi Channel State Information (CSI) samples with humans positioned at various locations, and training a machine learning model to infer the real-time location of intruders. Given the intensive human effort required for collecting sample data, we developed an algorithm to optimize the selection of locations for collecting CSI samples. The algorithm assesses each location’s contribution to the overall success rate of capturing intruders, thereby achieving optimal sample distribution. Our evaluation demonstrates that our approach achieves a capture rate of $\mathbf{7 8. 2 4 \%}$, which is up to $24 \%$ higher than baseline methods, despite being trained with data collected from only $13 \%$ of the locations.

Despite the widespread adoption of Web services in modern computing applications, there remains a lack of a systematic approach that can guide service developers in creating appealing services. This paper addresses this gap by presenting findings from a comprehensive study of RapidAPI web services, the largest service marketplace, and their integration into GitHub-hosted applications. We collected data on over 16K RapidAPI services and 19K corresponding GitHub repositories invoking these services, evaluating each service based on metrics such as latency, reliability, pricing, community support, and provider support. Our analysis examines how these metrics influence service popularity and usage patterns on GitHub. We manually analyzed 800 GitHub repositories and identified developers' service selection preferences and integration patterns, considering alternative services and their features. Additionally, we classified GitHub developers based on proficiency levels to understand how developers' levels of proficiency impact their service selection and integration strategies. Our findings offer insights for service marketplaces to recommend integration-friendly services and for service developers to create offerings tailored to real-world application needs.

Body Mass Index (BMI) is a critical metric for assessing public health and identifying populations at risk for obesity-related conditions. Traditional BMI monitoring methods often raise privacy concerns and require active cooperation from individuals, limiting their applicability in real-world scenarios. This paper introduces a novel approach to BMI monitoring that leverages proximal federated learning (PFL) using commodity WiFi devices. Our method addresses the challenges of data heterogeneity and intermittent connectivity in FL. By our approach, the Adaptive Elastic Stochastic Alternating Direction Method of Multipliers (AESADMM), an optimization algorithm designed to handle data heterogeneity and intermittent connectivity in FL scenarios, our system collects Channel State Information (CSI) from WiFi signals to passively classify BMI based on the impact of different body shapes on signal propagation. This approach ensures privacy preservation and eliminates the need for active participant involvement. Theoretical analysis and empirical results demonstrate the superior accuracy, reduced communication costs, and enhanced scalability of our proposed method compared to existing personalized FL frameworks, showcasing its potential as an effective tool for large-scale BMI monitoring in diverse environments.

MQTT is a widely utilized protocol in the IoT domain, specifically designed to minimize energy consumption in battery-powered, energy-intensive IoT devices. With the proliferation of smart home devices, there is a notable increase in co-located IoT devices capable of publishing to the same topic, as well as an increase in subscribers accessing diverse data from these devices. However, the architecture of current MQTT brokers do not effectively optimize task scheduling among multiple potential publishers. Our observations suggest that although aggregating sensing tasks on the same IoT device does not significantly impact the total sensing energy consumption, it substantially reduces communication energy costs by minimizing the long-tail energy expenses associated with wireless communications. In this paper, we introduce MQTT-EES (MQTT Energy Efficient Scheduling), which further optimizes the energy efficiency of MQTT by allocating sensing tasks on IoT devices with the goals of 1) minimizing long-tail communication energy consumption; and 2) prolonging the overall lifespan of the IoT system. We formulate the energy consumption challenge as an NP-hard problem and propose a greedy algorithm to tackle it. Our simulations show that MQTT-EES reduces average energy consumption by up to 12% and extends the overall lifespan of the system 2.78 times compared to standard MQTT implementations.

In recent years, the automotive industry and research communities have increasingly focused on exploring the potential of drones (also known as Unmanned Aerial Vehicles, or UAVs) to enhance ground vehicles. This exploration has opened a new avenue, termed "Drone-Car Collaboration", which presents exciting possibilities for joint operations. The collaboration between drones and cars promises increased situational awareness, expanded capabilities, enhanced driving safety, improved fuel efficiency, and novel applications. However, a detailed study is vital to understand the current progress and challenges in the drone-car domain. This paper presents the first comprehensive survey on drone-car collaboration, highlighting its many potential benefits. We explore state-of-the-art use cases and enabling technologies, and discuss future directions for realistically implementing drones in collaboration with ground vehicles.

ShakeAlert is an earthquake early warning (EEW) system that detects significant earthquakes so quickly that alerts can reach many people before shaking arrives. The current ShakeAlert system has been developed incrementally over the past decade, following an old-fashioned system design. The system's tight coupling between physical sensors and data processing modules has made it difficult to expand new sensors, adopt novel networking technology advances, or change the system's behavior at runtime, thus hindering the system's extensibility, scalability, and reliability. To address these limitations, this paper proposes to expand the existing system design by adding a "digital twin" middleware layer as virtual representations of physical sensor stations, which provides a standardized software interface for running distributed data processing applications and conceals the hardware/software differences in physical sensors. These virtual stations are placed at the edge of the network near physical stations, acting as a middleware layer between sensors and the ShakeAlert system. By incorporating digital twins, we can leverage cutting-edge networking technologies (such as edge/fog computing) to improve scalability, accept trustworthy sensor data from diverse sources to enhance extensibility, and move data processing functions closer to the sensor stations to reduce response time during large earthquakes when network throughput is impacted. Ultimately, this will enhance the reliability of the ShakeAlert system and help keep communities safe during earthquakes.

This paper explores the challenges of implementing Federated Learning (FL) in practical scenarios featuring isolated nodes with data heterogeneity, which can only be connected to the server through wireless links in an infrastructure-less environment. To overcome these challenges, we propose a novel mobilizing personalized FL approach, which aims to facilitate mobility and resilience. Specifically, we develop a novel optimization algorithm called Random Walk Stochastic Alternating Direction Method of Multipliers (RWSADMM). RWSADMM capitalizes on the server's random movement toward clients and formulates local proximity among their adjacent clients based on hard inequality constraints rather than requiring consensus updates or introducing bias via regularization methods. To mitigate the computational burden on the clients, an efficient stochastic solver of the approximated optimization problem is designed in RWSADMM, which provably converges to the stationary point almost surely in expectation. Our theoretical and empirical results demonstrate the provable fast convergence and substantial accuracy improvements achieved by RWSADMM compared to baseline methods, along with its benefits of reduced communication costs and enhanced scalability.

Numerous universities participate in student-led vehicle competitions, aiming to construct, examine, and race a student-built prototype vehicle, fostering learning and advancement in cutting-edge vehicle technologies. As electric vehicles and vehicle computing rise to prominence, modern vehicles have evolved into individualized computing hubs, equipped with sophisticated sensing, computing, and network capabilities. In response to these developments and future vehicle technology trends, there is a pressing need for roadside sensing systems in student vehicle competitions, facilitating student incorporation of emerging technologies to enhance vehicle safety and efficiency. However, acquisition of such systems poses financial hurdles. To solve this problem, this paper presents SensingBay, a ready-to-use and affordable roadside sensing system specifically designed for student vehicle competitions. SensingBay links sensor nodes via a WiFi-based mesh network to a central gateway node, which carries out more complex functions such as data analysis and user engagement. The system is constructed using Raspberry Pi, allowing the sensor nodes to be readily upgraded with a range of sensing capabilities. A prototype vehicle can interface with any sensor node to collect sensor data. The feasibility of SensingBay was validated by integrating laser distance sensors with the sensor node and employing the system in real-world conditions to collect real-time timing data for prototype vehicles. A performance evaluation of the system indicates that it has the potential to be a valuable resource in student vehicle competitions.

Widely used in modern software systems, web services have become a standard means of provisioning remote resources. As the number of available web services increases, multiple services that satisfy the same functional requirement can be used interchangeably. Given a set of interchangeable services, a software developer needs to find a web service that would provide the best performance and utility. However, web services are recommended based only on their system-related performance characteristics (so called QoS, whose properties include latency, reliability, availability, etc.), while their data-related performance characteristics (e.g., data freshness, correctness, coverage, etc.) are often overlooked. As a consequence, a recommended service may end up delivering information that is inaccurate or outdated, but with high performance. To address this problem, this paper introduces Quality of Information (QoI), a quality metric complementary to QoS, that measures to which degree a web service satisfies data-related non-functional requirements. To minimize the manual effort required to evaluate the results of invoking individual services, we introduce a comparative testing methodology based on the new concept of Objects of Interest (OI). By using OI, developers can normalize the relevant information obtained from dissimilar services, so it can be automatically compared. To concretely realize our ideas, we create QiSR, a system that recommends web services based on their QoI metrics. QiSR helps developers in determining how to match services’ input and output with application data requirements and how to measure the information quality of services. To evaluate the effectiveness of QiSR, we test it on representative manually selected web services. Our evaluation shows that services recommended based on both QoI and QoS exhibit better combined performance and utility than services recommend on QoS alone.

Project-based learning (PjBL) has been increasingly adopted in computer science courses to improve students’ engagement and learning outcomes. Although a computer networks course is in great need of a PjBL course module, no such module is available due to the huge gap between PjBL’s design requirements and the current structure and content of the course. This paper introduces a novel PjBL module for a computer networks course, which challenges the students with a real world problem of developing the communication system for a smart lock. Following the PjBL design principles, we devise several scaffolding activities and assignments, which can be integrated into a semester-long computer networks course. We test ran the PjBL module in both undergraduate- and graduate-level computer networks courses. Our preliminary evaluation results show that the proposed PjBL module is well received by the students and helps improve their learning outcomes.

To realize flexible networking and on-demand topology reconstructing, software-defined industrial networks (SDINs) are increasingly embracing the flat structure. Similar to software defined networks (SDN), SDIN suffers from low traffic scheduling efficiency caused by large and imbalanced flows, known as the heavy hitters problem. Due to such heavy hitters, industrial networks may fail to satisfy application’s QoS requirements, which results in more severe damages. To improve flow scheduling efficiency under heavy hitters, this article introduces a novel imitation learning-based flow scheduling (ILFS) method. ILFS utilizes P4-based In-band Network Telemetry (INT) to collect fine-grained, real-time traffic data from SDIN’s data plane. In the control plane, it integrates the Generative Adversarial Imitation Learning (GAIL) model with a soft actor critic to preserve the experiences of flow, thereby better scheduling large flows. Our experiments thoroughly compare ILFS’s performance with several state-of-the-art traffic scheduling strategies. The results indicate that ILFS successfully controls the link bandwidth the utilization between 10$\%$and 80$\%$and significantly improves the average network throughput and link utilization rate.

Mobile and energy harvesting devices increasingly provide resources for edge environments. These devices' mobility and limited energy budgets may cause failures and poor performance. The reliability and efficiency of edge services can be improved with equivalent microservices that satisfy application requirements by different means: execute equivalent microservices in the predefined patterns of fail-over to minimize execution costs or speculative parallelism to reduce latency. However, given the vast dissimilarities in resource availability and capability across edge environments, being limited to these predefined patterns when implementing edge services causes inconsistent QoS. To address this problem, we provide QoS-consistent edge services by customizing the execution of equivalent microservices. Our system estimates the environment-specific QoS of equivalent microservices and dynamically generates execution strategies that best satisfy given QoS requirements. We evaluate the effectiveness and performance of our system via simulations and benchmarks with realistic edge deployments. Our approach consistently out-performs the predefined execution patterns in satisfying the QoS requirements in unreliable and dynamic edge environments.

Participatory sensing uses both local devices for data collection and cloud-based servers for processing. However, transferring the collected data to the cloud can lead to draining device battery power and cause network bandwidth bottlenecks, especially for large multimedia files. In this paper, we investigate how the processing resources at the edge of the network can be leveraged to enable efficient participatory sensing that avoids heavy network traffic. In particular, we report on the experiences of designing, implementing, and evaluating a sensing system that constructs indoor maps by recognizing door signs. A distinguishing characteristic of our system is an almost exclusive use of edge-based processing for tasks that include ML-based image recognition, human-assisted data verification, data model retraining, and administrative data flow aggregation. Our evaluation shows that our system architecture effectively leverages the available edge resources, while greatly reducing network traffic. Based on our experiences of implementing and evaluating our system prototype, we identify several open research directions for further advancing edge-based participatory sensing.

Equivalent services deliver the same functionality with dissimilar non-functional characteristics, including latency, accuracy, and cost. With these dissimilarities in mind, developers can exploit the combined execution of equivalent services to increase accuracy, shorten latency, or reduce cost. However, it remains unknown how to effectively combine equivalent services to satisfy application requirements. With the recent surge in popularity of machine learning, different vendors offer a plethora of equivalent services, whose characteristics are mostly undocumented. As a result, developers cannot make an informed decision about which service to select from a set of equivalent services. To address this problem, we explore different service combination strategies (i.e., majority voting, weighted-majority voting, stacking, and custom) to ascertain their impact on non-functional characteristics. In particular, we study how these strategies impact the accuracy, cost, and latency of the face detection task and validate our findings on the sentiment analysis task. We consider the combined executions of commercial web services, deployed in the cloud, and open-source implementations, deployed as edge services. Our evaluation reveals that the combined execution of equivalent services is most effective for improving cost and latency. Informed by our experimental results, we formulate practical guidelines to help developers identify the best execution strategy for a given set of services.

The applicability of the microservice architecture has extended beyond traditional web services, making steady inroads into the domains of IoT and edge computing. Due to dissimilar contexts in different execution environments and inherent mobility, edge and IoT applications suffer from low execution reliability. Replication, traditionally used to increase service reliability and scalability, is inapplicable in these resource-scarce environments. Alternately, programmers can orchestrate the parallel or sequential execution of equivalent microservices-microservices that provide the same functionality by different means. Unfortunately, the resulting orchestrations rely on parallelization, synchronization, and failure handing, all tedious and error-prone to implement. Although automated orchestration shifts the burden of generating workflows from the programmer to the compiler, existing programming models lack both syntactic and semantic support for equivalence. In this paper, we enhance compiler-generated execution orchestration with equivalence to efficiently increase reliability. We introduce a dataflow-based domain-specific language, whose dataflow specifications include the implicit declarations of equivalent microservices and their execution patterns. To automatically generate reliable workflows and execute them efficiently, we introduce new equivalence workflow constructs. Our evaluation results indicate that our solution can effectively and efficiently increase the reliability of microservice-based applications.

Modern mobile devices feature ever increasing computational, sensory, and network resources, which can be shared to execute tasks on behalf of nearby devices. Mobile device clouds (MDCs) facilitate such distributed execution by exposing the collective resources of a set of nearby mobile devices through a unified programming interface. However, the true potential of MDCs remains untapped, as they fail to provide practical programming support for developers to execute distributed functionalities. To address this problem, we introduce a microservice-based Programmable MDC architecture (PMDC), highly customized for the unique features of MDC environments. PMDC conveniently provisions functionalities as microservices, which are deployed on MDC devices on demand. PMDC features a novel domain specific language that provides abstractions for concisely expressing fine-grained control over the procedures of device capability sharing and microservice execution. Furthermore, PMDC introduces a new system component-the microservice gateway, which reconciles the supply of available device capabilities and the demand for microservice execution to distribute microservices within an MDC. Our evaluation shows that MDCs, expressed by developers through the PMDC declarative programming interface, exhibit low energy consumption and high performance.

Mobile applications and Internet of Things applications increasingly require one mobile device to exchange data with another collocated device. Despite the introduction of multiple standard device-to-device communication protocols (i.e., WiFi direct, Bluetooth, Bluetooth Low Energy, NFC), transferring data across devices remains hard for mobile programmers for three reasons. 1) different devices support different D2D communication interfaces, and therefore the available D2D communication channels are dynamically decided by the peers; 2) different D2D interfaces have different features (connection establish time, data rate, energy consumption) and different performances under different contexts, complicating the decision which channel to use; 3) implementing and debugging data transmission functionalities requires knowing the low-level details of communication protocols, which is difficult and error prone. To solve the above mentioned problem, this paper presents a runtime framework for context-sensitive device-to-device communication. The presented framework consists of two major components: 1) a set of encapsulated D2D data transmission interfaces to reduce the programming efforts; 2) a context-sensitive communication channel selection algorithm to select the optimal communication channel as defined by the dynamic context. We design and implement the runtime framework, and evaluate its performance in terms of the required programming effort, energy consumption and data transmission latency under different contexts.

Compartmental models have been used to model information diffusion on social media. However, there have been few studies on modelling positive and negative public opinions using compartmental models. This study aimed for using sentiment analysis and compartmental model to model the propagation of positive and negative opinions on microblogging big media. The authors studied the news propagation of seven popular social topics on China's Sina Weibo microblogging platform. Natural language processing and sentiment analysis were used to identify public opinions from microblogging big data. Then two existing (SIZ and SEIZ) models and a newly developed (SE2IZ) model were implemented to model the news propagation and evaluate the trends of public opinions on selected social topics. Simulation study was used to check model fitting performance. The results show that the new SE2IZ model has a better model fitting performance than existing models. This study sheds some new light on using social media for public opinion estimation and prediction.

One of the fundamental building blocks of a mobile application is the ability to persist program data between different invocations. Referred to as persistence, this functionality is commonly implemented by means of persistence frameworks. When choosing a particular framework, Android—the most popular mobile platform—offers a wide variety of options to developers. Unfortunately, the energy, performance, and programming effort trade-offs of these frameworks are poorly understood, leaving the Android developer in the dark trying to select the most appropriate option for their applications. To address this problem, this paper reports on the results of the first systematic study of six Android persistence frameworks (i.e., ActiveAndroid, greenDAO, OrmLite, Sugar ORM, Android SQLite, and Java Realm) in their application to and performance with popular benchmarks, such as DaCapo. Having measured and analyzed the energy, performance, and programming effort trade-offs for each framework, we present a set of practical guidelines for the developer to choose between Android persistence frameworks. Our findings can also help the framework developers to optimize their products to meet the desired design objectives.

Location based services are getting increasingly popular in participatory sensing systems. They make use of location information on the mobile devices to support applications that improve personal health, object search, and entertainment. However, GPS positioning consumes a lot of energy, which can drain a mobile device's battery. Although WiFi localization and cell tower localization have been suggested as alternatives, they have lower localization accuracy and limited coverage. In this paper, we suggest a novel solution for multiple mobile devices to perform collaborative localization to reduce energy consumption and provide accurate localization. We divide the mobile devices into two groups, the aggregator group and the collector group. The aggregator group turns on their GPS periodically, while the collector group uses the locations of the aggregators to estimate their own locations. We formulate the aggregator set selection problem and propose two novel algorithms to minimize the energy consumption in collaborative localization. Simulations with real traces showed that our proposed solution can save up to 88% of the energy of the entire network.

The increasingly popular smartphones enable participatory sensing systems to collect location-based sensing data for different tasks. However, GPS positioning is very energy consuming, which could drain a mobile device's battery quickly. High energy consumption may threaten the participants and reduce the sustainability of the participatory sensing systems. In this paper, we propose a collaborative localization strategy with load balancing. Simulations with real traces showed that our proposed solution can save more than 80% of the energy consumption for localization in the entire network with load balancing.

Participatory sensing systems can be used for concurrent event monitoring applications, like noise levels, fire, and pollutant concentrations. However, they are facing new challenges as to how to accurately detect the exact boundaries of these events, and further, to select the most appropriate participants to collect the sensing data. On the one hand, participants’ handheld smart devices are constrained with different energy conditions and sensing capabilities, and they move around with uncontrollable mobility patterns in their daily life. On the other hand, these sensing tasks are within time-varying quality-of-information (QoI) requirements and budget to afford the users’ incentive expectations. Toward this end, this article proposes an event-driven QoI-aware participatory sensing framework with energy and budget constraints. The main method of this framework is event boundary detection. For the former, a two-step heuristic solution is proposed where the coarse-grained detection step finds its approximation and the fine-grained detection step identifies the exact location. Participants are selected by explicitly considering their mobility pattern, required QoI of multiple tasks, and users’ incentive requirements, under the constraint of an aggregated task budget. Extensive experimental results, based on a real trace in Beijing, show the effectiveness and robustness of our approach, while comparing with existing schemes.

Incentive allocation is an important research issue in participatory sensing as it determines the willingness of participants in joining the sensing campaign. Existing incentive approaches either decide the payments without consulting the participants, or require burdensome negotiation procedures like bid-price auction. In this paper, we propose a novel incentive allocation mechanism, which encourages participation and allocates incentives dynamically to achieve accurate sensing results. The proposed mechanism consists of two major elements. The first is a lightweight incentive negotiation procedure, which dynamically offers incentives to participants in spatio-temporal subregions and collects their responses. The second is the optimization problem for incentive allocation as well as it's solution, which aims at maximizing data quality by capturing the amount and the distribution of data samples. Simulations with real datasets confirmed that the proposed solution can provide dynamic incentive offers according to the estimated value of participants' data contribution to the overall quality of sensing result.

In increasingly popular participatory sensing systems, new challenges are arising to select the most appropriate participants when considering their hand-held smart device's different energy conditions, uncontrollable mobility pattern, and associated sensing capabilities to best satisfy the quality-of-information (QoI) requirements of sensing tasks. This paper proposes a QoI-aware energy-efficient participant selection strategy, where four key design elements are proposed. First is QoI satisfaction metric of a sensing task that uses the data granularity and quantity collected by participants to measure to what extend the task's QoI requirements are satisfied. Second is an “energy consumption index”, which estimates the impact of energy cost during the data collection on different participant's smart devices with different remaining energy levels. Third is the estimation of the collected amount of data by participants, where a probability-based movement model is proposed. Fourth is the proposal of a multi-objective constrained optimization problem for participant selection, where task QoI requirements and energy consumption index of all participants are taken as optimization objectives, and solved by our proposed suboptimal, easy-to-implement solution. Real and extensive trace-based experiments show that, the proposed participant selection scheme can well balance the trade-off between the task QoI and energy consumptions by selecting most efficient participants, compared with existing schemes.

In some practical scenarios such as tour guiding and children babysitting, one mobile device held by tour guides or parents need to know the distance and direction of another nearby mobile device held by tourist or children. However, to date, most existing pedestrian localization methods rely on a fixed external infrastructure, such as a global positioning system(GPS) or pre-deployed wifi access points to provide Localization service for mobile devices. Such methods are constrained either by limited GPS coverage or by complicated set-up procedures. We observe that, when two devices are moving, the change of their positions leads to the change of distance between them. Given the same movements, different relative locations between devices lead to different distance changes. Besides, the distance between and relative movement of devices can be measured by two phone-embedded sensors respectively. This motivates us to exploit the relative localization method by merely two mobile devices. In this work, we present Phone-Radar, which is an infrastructure-free device-to-device localization system. According to the propagation model of wireless signals, the change of distance between devices are modeled by the change of wireless signal strength between them. The movements of devices are recorded by the inertial sensors using step-counting method. We further study the relationship among the initial relative locations between the two devices, their relative movements and the change of received signal strength measurements. Moreover, we implement the proposed method and measure its performance under real world conditions. The testbed experiments show the efficiency of our proposed method.

Incentive strategy is important in participatory sensing, especially when the budget is limited, to decide how much and where the samples should be collected. Current auction-based incentive strategies purchase sensing data with lowest price requirements to maximize the amount of samples. However, such methods may lead to inaccurate sensing result after data interpolation, particularly for participants that are massing in certain subregions where the low-price sensing data are usually aggregated. In this paper, we introduce weighted entropy as a quantitative metric to evaluate the distribution of samples and find that the distribution of data samples is another important factor to the accuracy of sensing result. We further propose a greedy-based incentive strategy which considers both the amount and distribution of samples in data collection. Simulations with real datasets confirmed the impact of samples distribution to data accuracy and demonstrated the efficacy of our proposed incentive strategy.