Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00020
Tom Goethals, Merlijn Sebrechts, M. Al-Naday, B. Volckaert, F. Turck
The recent rise of edge computing and FaaS triggered a revolution in the field of software virtualization, improving performance and security. This paper benchmarks various lightweight virtualization platforms, such as microVMs and containers, in the context of edge microservices. Factors taken into account include toolchain maturity, networking capabilities, boot time, resource use, microservice performance, and ARM architecture readiness. We present a functional comparison and benchmarks on both a Raspberry Pi 4 and an x86-64 platform. The results indicate standard Docker containers offer reliable performance and low memory use, while microVM-based solutions such as Firecracker are more isolated. Moreover, OSv unikernels have extremely low boot times and significantly better performance than Docker containers. Finally, while gVisor offers improved security and excellent compatibility, its performance is only 10% of default Docker performance.
最近边缘计算和FaaS的兴起引发了软件虚拟化领域的一场革命，提高了性能和安全性。本文在边缘微服务环境中对各种轻量级虚拟化平台(如微虚拟机和容器)进行了基准测试。考虑的因素包括工具链成熟度、网络功能、引导时间、资源使用、微服务性能和ARM架构准备情况。我们在Raspberry Pi 4和x86-64平台上进行了功能比较和基准测试。结果表明，标准Docker容器提供可靠的性能和较低的内存使用，而基于microvm的解决方案(如爆竹)则更加孤立。此外，OSv内核的启动时间极短，性能明显优于Docker容器。最后，虽然gVisor提供了改进的安全性和出色的兼容性，但它的性能只有默认Docker性能的10%。
{"title":"A Functional and Performance Benchmark of Lightweight Virtualization Platforms for Edge Computing","authors":"Tom Goethals, Merlijn Sebrechts, M. Al-Naday, B. Volckaert, F. Turck","doi":"10.1109/EDGE55608.2022.00020","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00020","url":null,"abstract":"The recent rise of edge computing and FaaS triggered a revolution in the field of software virtualization, improving performance and security. This paper benchmarks various lightweight virtualization platforms, such as microVMs and containers, in the context of edge microservices. Factors taken into account include toolchain maturity, networking capabilities, boot time, resource use, microservice performance, and ARM architecture readiness. We present a functional comparison and benchmarks on both a Raspberry Pi 4 and an x86-64 platform. The results indicate standard Docker containers offer reliable performance and low memory use, while microVM-based solutions such as Firecracker are more isolated. Moreover, OSv unikernels have extremely low boot times and significantly better performance than Docker containers. Finally, while gVisor offers improved security and excellent compatibility, its performance is only 10% of default Docker performance.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125399814","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00035
Roman Karlstetter, R. Widhopf-Fenk, Martin Schulz
Sensor data is of crucial importance in many IoT scenarios. It is used for online monitoring as well as long term data analytics, enabling countless use cases from damage prevention to predictive maintenance. Multivariate sensor time series data is acquired and initially stored close to the sensor, at the edge. It is also beneficial to summarize this data in windowed aggregations at different resolutions. A subset of the resulting aggregation hierarchy is typically sent to a cloud infrastructure, often via intermittent or low bandwidth connections. Consequently, different views on the data exist on different nodes in the edge-to-cloud continuum. However, when querying this data, users are interested in a fast response and a complete, unified view on the data, regardless of which part in the infrastructure continuum they send the query to and where the data is physically stored. In this paper, we present a loosely coupled approach that enables fast range queries on a distributed and hierarchical sensor database. Our system only assumes the possibility of fast local range queries on a hierarchical sensor database. It does not require any shared state between nodes and thus degrades gracefully in case certain parts of the hierarchy are unreachable. We show that our system is suitable for driving interactive data exploration sessions on terabytes of data while unifying the different views on the data. Thus, our system can improve the data analysis experience in many geo-distributed scenarios.
{"title":"Querying Distributed Sensor Streams in the Edge-to-Cloud Continuum","authors":"Roman Karlstetter, R. Widhopf-Fenk, Martin Schulz","doi":"10.1109/EDGE55608.2022.00035","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00035","url":null,"abstract":"Sensor data is of crucial importance in many IoT scenarios. It is used for online monitoring as well as long term data analytics, enabling countless use cases from damage prevention to predictive maintenance. Multivariate sensor time series data is acquired and initially stored close to the sensor, at the edge. It is also beneficial to summarize this data in windowed aggregations at different resolutions. A subset of the resulting aggregation hierarchy is typically sent to a cloud infrastructure, often via intermittent or low bandwidth connections. Consequently, different views on the data exist on different nodes in the edge-to-cloud continuum. However, when querying this data, users are interested in a fast response and a complete, unified view on the data, regardless of which part in the infrastructure continuum they send the query to and where the data is physically stored. In this paper, we present a loosely coupled approach that enables fast range queries on a distributed and hierarchical sensor database. Our system only assumes the possibility of fast local range queries on a hierarchical sensor database. It does not require any shared state between nodes and thus degrades gracefully in case certain parts of the hierarchy are unreachable. We show that our system is suitable for driving interactive data exploration sessions on terabytes of data while unifying the different views on the data. Thus, our system can improve the data analysis experience in many geo-distributed scenarios.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132841226","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00027
Chengzu Dong, Frank Jiang, Shiping Chen, Xiao Liu
In recent years, Unmanned Aerial Vehicles (UAVs) have been widely used in smart logistics to address the last-mile delivery challenge. To complete the delivery tasks, UAVs usually need to store or have access to some sensitive data such as the private information about the parcel and its receiver during their flight, and only the authenticated people or devices are authorised to access the UAV's data. However, there are many potential risks such as injection attacks and spoofing attacks that can cause data leakage. Therefore, how to preserve users' sensitive data and authenticate the identities of people and devices in the UAV delivery system has become an essential requirement under the emerging zero-trust security framework. In this paper, to address such an issue, we propose a continuous authentication scheme for UAV delivery systems that is featured with continuous authentication, multi-factor authentication (MFA), and digital authentication using the blockchain-based Self-Sovereign Identity (SSI) technique. Specifically, continuous authentication is applied to both users and UAVs in the system. The evaluation results demonstrate that our proposed continuous authentication scheme can effectively prevent various attacks and provide security analysis for UAV delivery systems.
{"title":"Continuous Authentication for UAV Delivery Systems Under Zero-Trust Security Framework","authors":"Chengzu Dong, Frank Jiang, Shiping Chen, Xiao Liu","doi":"10.1109/EDGE55608.2022.00027","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00027","url":null,"abstract":"In recent years, Unmanned Aerial Vehicles (UAVs) have been widely used in smart logistics to address the last-mile delivery challenge. To complete the delivery tasks, UAVs usually need to store or have access to some sensitive data such as the private information about the parcel and its receiver during their flight, and only the authenticated people or devices are authorised to access the UAV's data. However, there are many potential risks such as injection attacks and spoofing attacks that can cause data leakage. Therefore, how to preserve users' sensitive data and authenticate the identities of people and devices in the UAV delivery system has become an essential requirement under the emerging zero-trust security framework. In this paper, to address such an issue, we propose a continuous authentication scheme for UAV delivery systems that is featured with continuous authentication, multi-factor authentication (MFA), and digital authentication using the blockchain-based Self-Sovereign Identity (SSI) technique. Specifically, continuous authentication is applied to both users and UAVs in the system. The evaluation results demonstrate that our proposed continuous authentication scheme can effectively prevent various attacks and provide security analysis for UAV delivery systems.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124145910","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00024
Kaustubh Rajendra Rajput, Chinmay Kulkarni, Byungjin Cho, W. Wang, I. Kim
Due to the development of small-size, energy-efficient, and powerful CPUs and GPUs for single board computers, various edge devices are widely adopted for hosting real-world applications, including real-time object detection, autonomous driving, and sensor stream processing. At the same time, serverless computing receives increasing attention as a new application deployment model because of its simplicity, scalability, event-driven processing, and short-lived computation. Therefore, there is a growing demand for applying serverless computing to edge computing environments. However, due to the lack of characterization of serverless edge computing (e.g., application performance and impact from resource heterogeneity), researchers and practitioners have to conduct tedious measurements to understand the performance of serverless applications on edge devices in non-systematic ways.We create EdgeFaaSBench, a novel benchmark suite for serverless computing on edge devices, to bridge this gap. EdgeFaaSBench is developed on top of Apache OpenFaaS with Docker Swarm and can run various serverless benchmark workloads on edge devices with different hardware specifications (e.g., GPUs). EdgeFaaSBench contains 14 different benchmark workloads running on heterogeneous edge devices and captures various system-level, application-level, and serverless-specific metrics, including system utilization, response time, cold/warm start times, and impact of concurrent function executions. Experimental studies are conducted on two widely used edge devices, Raspberry Pi 4B and Jetson Nano, to show EdgeFaaSBench’s capabilities to benchmark serverless computing on edge devices.
{"title":"EdgeFaaSBench: Benchmarking Edge Devices Using Serverless Computing","authors":"Kaustubh Rajendra Rajput, Chinmay Kulkarni, Byungjin Cho, W. Wang, I. Kim","doi":"10.1109/EDGE55608.2022.00024","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00024","url":null,"abstract":"Due to the development of small-size, energy-efficient, and powerful CPUs and GPUs for single board computers, various edge devices are widely adopted for hosting real-world applications, including real-time object detection, autonomous driving, and sensor stream processing. At the same time, serverless computing receives increasing attention as a new application deployment model because of its simplicity, scalability, event-driven processing, and short-lived computation. Therefore, there is a growing demand for applying serverless computing to edge computing environments. However, due to the lack of characterization of serverless edge computing (e.g., application performance and impact from resource heterogeneity), researchers and practitioners have to conduct tedious measurements to understand the performance of serverless applications on edge devices in non-systematic ways.We create EdgeFaaSBench, a novel benchmark suite for serverless computing on edge devices, to bridge this gap. EdgeFaaSBench is developed on top of Apache OpenFaaS with Docker Swarm and can run various serverless benchmark workloads on edge devices with different hardware specifications (e.g., GPUs). EdgeFaaSBench contains 14 different benchmark workloads running on heterogeneous edge devices and captures various system-level, application-level, and serverless-specific metrics, including system utilization, response time, cold/warm start times, and impact of concurrent function executions. Experimental studies are conducted on two widely used edge devices, Raspberry Pi 4B and Jetson Nano, to show EdgeFaaSBench’s capabilities to benchmark serverless computing on edge devices.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114607180","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00023
Arthur Desuert, Stéphanie Chollet, Laurent Pion, D. Hély
Connected devices are being deployed at a steady rate, providing services like data collection. Pervasive applications rely on those edge devices to seamlessly provide services to users. To connect applications and edge devices, using a middleware has been a popular approach. The research is active on the subject as there are many open challenges. The secure management of the edge devices and the security of the middleware are two of them. As security is a crucial requirement for pervasive environment, we propose a middleware architecture easing the secure use of edge devices for pervasive applications, while supporting the heterogeneity of communication protocols and the dynamism of devices. Because of the heterogeneity in protocols and security features, not all edge devices are equally secure. To allow the pervasive applications to gain control over this heterogeneous security, we propose a model to describe edge devices security. This model is accessible by the applications through our middleware. To validate our work, we developed a demonstrator of our middleware and we tested it in a concrete scenario.
{"title":"A Middleware for Secure Integration of Heterogeneous Edge Devices","authors":"Arthur Desuert, Stéphanie Chollet, Laurent Pion, D. Hély","doi":"10.1109/EDGE55608.2022.00023","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00023","url":null,"abstract":"Connected devices are being deployed at a steady rate, providing services like data collection. Pervasive applications rely on those edge devices to seamlessly provide services to users. To connect applications and edge devices, using a middleware has been a popular approach. The research is active on the subject as there are many open challenges. The secure management of the edge devices and the security of the middleware are two of them. As security is a crucial requirement for pervasive environment, we propose a middleware architecture easing the secure use of edge devices for pervasive applications, while supporting the heterogeneity of communication protocols and the dynamism of devices. Because of the heterogeneity in protocols and security features, not all edge devices are equally secure. To allow the pervasive applications to gain control over this heterogeneous security, we propose a model to describe edge devices security. This model is accessible by the applications through our middleware. To validate our work, we developed a demonstrator of our middleware and we tested it in a concrete scenario.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115270310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00015
Zahidur Talukder, M. A. Islam
Federated Learning (FL) offers a privacy-preserving massively distributed Machine Learning (ML) paradigm where many clients cooperatively work together towards training a shared machine learning model. FL, however, is susceptible to data heterogeneity problems as the FL clients have diverse data sources. Prior works employ auto-weighted model aggregation to mitigate the heterogeneity issue to minimize the impact of unfavorable model updates. However, existing approaches require extensive computation for statistical analysis of clients’ model updates. To circumvent this, we propose, FedASL (Federated Learning with Auto-weighted Aggregation based on Standard Deviation of Training Loss) which uses only the local training loss of FL clients for auto-weighting the model aggregation. Our evaluation under three different datasets and various data corruption scenarios reveals that FedASL can effectively thwart data corruption from bad clients while causing as little as one-tenth of the computation cost of existing approaches.
{"title":"Computationally Efficient Auto-Weighted Aggregation for Heterogeneous Federated Learning","authors":"Zahidur Talukder, M. A. Islam","doi":"10.1109/EDGE55608.2022.00015","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00015","url":null,"abstract":"Federated Learning (FL) offers a privacy-preserving massively distributed Machine Learning (ML) paradigm where many clients cooperatively work together towards training a shared machine learning model. FL, however, is susceptible to data heterogeneity problems as the FL clients have diverse data sources. Prior works employ auto-weighted model aggregation to mitigate the heterogeneity issue to minimize the impact of unfavorable model updates. However, existing approaches require extensive computation for statistical analysis of clients’ model updates. To circumvent this, we propose, FedASL (Federated Learning with Auto-weighted Aggregation based on Standard Deviation of Training Loss) which uses only the local training loss of FL clients for auto-weighting the model aggregation. Our evaluation under three different datasets and various data corruption scenarios reveals that FedASL can effectively thwart data corruption from bad clients while causing as little as one-tenth of the computation cost of existing approaches.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125186052","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00026
A. Telikani, Jun Shen, Peng Wang, Jie Yang
Due to the wide adoption and deployment of the Internet of Things (IoT), massive amounts of data are being generated and shared across various sectors. Privacy disclosure is a major threat in IoT-related applications if collected data is directly outsourced. In IoT environments with large datasets, the existing Privacy-Preserving Data Mining (PPDM) mechanisms are inefficient and not scalable. To deal with this shortcoming, we develop a novel evolutionary PPDM framework, namely GPU-Enabled PPDM for IoT (GEPI), using GPUs at the edge layer to make the PPDM both efficient and exhibiting usefulness for IoT applications. On the one hand, the evolutionary algorithm used in the PPDM can guarantee the high utility by selecting the best candidate transactions for modification, providing a shareable dataset with minimum modifications and maximum privacy. On the other hand, the evolutionary algorithm is parallelized using a developed GPU based mechanism to accelerate database scans. In our mechanism, the dataset is distributed among the GPU threads to compute the fitness value in a parallel manner. Tests with extensive benchmarks reveal that our mechanism can accelerate the fitness function step 53.7x on average. The findings also show that the developed PPDM algorithm achieves an average speedup of 43.8x and 47.3x when compared to the state-of-the-art algorithms of ABC4ARH and PACO2DT, respectively.
由于物联网(IoT)的广泛采用和部署，正在生成大量数据并在各个部门之间共享。如果收集的数据被直接外包，隐私泄露将是物联网相关应用的主要威胁。在具有大数据集的物联网环境中，现有的隐私保护数据挖掘(PPDM)机制效率低下且不可扩展。为了解决这一缺点，我们开发了一种新的渐进式PPDM框架，即GPU-Enabled PPDM for IoT (GEPI)，在边缘层使用gpu，使PPDM既高效又适用于物联网应用。一方面，PPDM中使用的进化算法通过选择最优的候选事务进行修改，保证了较高的效用，提供了一个修改最少、隐私最大的可共享数据集。另一方面，进化算法使用开发的基于GPU的机制并行化，以加速数据库扫描。在我们的机制中，数据集分布在GPU线程之间，以并行方式计算适应度值。大量的基准测试表明，我们的机制可以平均加速适应度函数步长53.7倍。研究结果还表明，与最先进的ABC4ARH和PACO2DT算法相比，所开发的PPDM算法分别实现了43.8倍和47.3倍的平均加速。
{"title":"A privacy-aware data sharing framework for Internet of Things through edge computing platform","authors":"A. Telikani, Jun Shen, Peng Wang, Jie Yang","doi":"10.1109/EDGE55608.2022.00026","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00026","url":null,"abstract":"Due to the wide adoption and deployment of the Internet of Things (IoT), massive amounts of data are being generated and shared across various sectors. Privacy disclosure is a major threat in IoT-related applications if collected data is directly outsourced. In IoT environments with large datasets, the existing Privacy-Preserving Data Mining (PPDM) mechanisms are inefficient and not scalable. To deal with this shortcoming, we develop a novel evolutionary PPDM framework, namely GPU-Enabled PPDM for IoT (GEPI), using GPUs at the edge layer to make the PPDM both efficient and exhibiting usefulness for IoT applications. On the one hand, the evolutionary algorithm used in the PPDM can guarantee the high utility by selecting the best candidate transactions for modification, providing a shareable dataset with minimum modifications and maximum privacy. On the other hand, the evolutionary algorithm is parallelized using a developed GPU based mechanism to accelerate database scans. In our mechanism, the dataset is distributed among the GPU threads to compute the fitness value in a parallel manner. Tests with extensive benchmarks reveal that our mechanism can accelerate the fitness function step 53.7x on average. The findings also show that the developed PPDM algorithm achieves an average speedup of 43.8x and 47.3x when compared to the state-of-the-art algorithms of ABC4ARH and PACO2DT, respectively.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134383762","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00031
Dixit Bhatta, Lena Mashayekhy
IoT (Internet-of-Things) devices require both reliable, ultra-low latency connection and on-demand access to computing resources in their vicinity. Edge computing can provide nearby computing resources through offloading to cloudlets. However, challenges lie in providing stable services in big cities, emergency situations, and places lacking sufficient infrastructure due to dynamic IoT demands and their uncertain mobility, coupled with intermittent or limited connectivity to nearby computing resources. There has been an increasing adoption of UAV-mounted mobile cloudlets to address these issues, where mobile cloudlets can be seamlessly integrated into the existing network infrastructure for offloading and enhanced connectivity. However, adaptively placing mobile cloudlets under continuously evolving conditions is still an open problem. In this paper, we study mobile cloudlet placement in highly dynamic, next-generation edge networks. Our objective is to maintain stable ultra-low-latency services and enhance coverage by optimizing the placement of mobile cloudlets in highly dynamic scenarios. We propose a novel approach for this problem using inspirations from concepts in physics, namely the center of gravity and force of attraction. Our approach emulates how physical bodies maintain balance and adjust their positions when acted upon by dynamic external forces for efficient placement of mobile cloudlets in response to dynamic and uncertain device mobility. The results show that our proposed approach provides high device coverage, excellent energy efficiency, and stable edge services in real-time.
{"title":"Physics-Inspired Mobile Cloudlet Placement in Next-Generation Edge Networks","authors":"Dixit Bhatta, Lena Mashayekhy","doi":"10.1109/EDGE55608.2022.00031","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00031","url":null,"abstract":"IoT (Internet-of-Things) devices require both reliable, ultra-low latency connection and on-demand access to computing resources in their vicinity. Edge computing can provide nearby computing resources through offloading to cloudlets. However, challenges lie in providing stable services in big cities, emergency situations, and places lacking sufficient infrastructure due to dynamic IoT demands and their uncertain mobility, coupled with intermittent or limited connectivity to nearby computing resources. There has been an increasing adoption of UAV-mounted mobile cloudlets to address these issues, where mobile cloudlets can be seamlessly integrated into the existing network infrastructure for offloading and enhanced connectivity. However, adaptively placing mobile cloudlets under continuously evolving conditions is still an open problem. In this paper, we study mobile cloudlet placement in highly dynamic, next-generation edge networks. Our objective is to maintain stable ultra-low-latency services and enhance coverage by optimizing the placement of mobile cloudlets in highly dynamic scenarios. We propose a novel approach for this problem using inspirations from concepts in physics, namely the center of gravity and force of attraction. Our approach emulates how physical bodies maintain balance and adjust their positions when acted upon by dynamic external forces for efficient placement of mobile cloudlets in response to dynamic and uncertain device mobility. The results show that our proposed approach provides high device coverage, excellent energy efficiency, and stable edge services in real-time.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"541 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123578574","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00032
Fouad Sakr, Riccardo Berta, Joseph Doyle, H. Younes, A. De Gloria, F. Bellotti
Binarization is a machine learning optimization for limited resource devices that has achieved significant results in edge applications. As microcontrollers are the mainstream platform for field applications in industry, this article investigates memory efficient deployment of binary neural networks (BNNs) to such devices. To this end, we developed a library of binarized NN layers in platform-independent C language. The implementation optimizes the memory footprint by adapting the number of filters and neurons for each layer to the C language specifications (8 bit addressable spacing and 32 bit floating points). We tested the library by implementing a BNN for robotic touch classification on a Cortex-M7 microcontroller. Experimental results show the validity of the proposed hardware and software architecture, with improvements over the state of the art in robotic touch classification in terms of accuracy (+2%), memory footprint (+15%), and latency (+25%).
{"title":"Memory Efficient Binary Convolutional Neural Networks on Microcontrollers","authors":"Fouad Sakr, Riccardo Berta, Joseph Doyle, H. Younes, A. De Gloria, F. Bellotti","doi":"10.1109/EDGE55608.2022.00032","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00032","url":null,"abstract":"Binarization is a machine learning optimization for limited resource devices that has achieved significant results in edge applications. As microcontrollers are the mainstream platform for field applications in industry, this article investigates memory efficient deployment of binary neural networks (BNNs) to such devices. To this end, we developed a library of binarized NN layers in platform-independent C language. The implementation optimizes the memory footprint by adapting the number of filters and neurons for each layer to the C language specifications (8 bit addressable spacing and 32 bit floating points). We tested the library by implementing a BNN for robotic touch classification on a Cortex-M7 microcontroller. Experimental results show the validity of the proposed hardware and software architecture, with improvements over the state of the art in robotic touch classification in terms of accuracy (+2%), memory footprint (+15%), and latency (+25%).","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131195435","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-07-01DOI: 10.1109/EDGE55608.2022.00017
C. Dehury, Praveen Kumar Donta, S. Dustdar, S. Srirama
The recent advancement in edge computing allows IoT devices to be smart enough to collect the surrounding data and react to the environment based on a predefined logic set or the instructions from a remotely located control center such as cloud computing. However, the intelligence in this smart environment is primarily implemented in the central controller, allowing no or little room for IoT devices to collaborate, share knowledge, and take advantage of peer devices’ knowledge with zero or minimal reliance on the central controller. This leaves the whole system not scalable and inefficient in energy consumption. In this paper, we propose an approach where edge intelligence is imposed in a clustered and cohesive manner. This enables the IoT devices to form one or more clusters based on the intelligence they possess and operate collaboratively with or without any intervention of the remote control center. This allows IoT devices to efficiently collect and react to the environment, resulting in better service quality and further reducing overall energy consumption.
{"title":"CCEI-IoT: Clustered and Cohesive Edge Intelligence in Internet of Things","authors":"C. Dehury, Praveen Kumar Donta, S. Dustdar, S. Srirama","doi":"10.1109/EDGE55608.2022.00017","DOIUrl":"https://doi.org/10.1109/EDGE55608.2022.00017","url":null,"abstract":"The recent advancement in edge computing allows IoT devices to be smart enough to collect the surrounding data and react to the environment based on a predefined logic set or the instructions from a remotely located control center such as cloud computing. However, the intelligence in this smart environment is primarily implemented in the central controller, allowing no or little room for IoT devices to collaborate, share knowledge, and take advantage of peer devices’ knowledge with zero or minimal reliance on the central controller. This leaves the whole system not scalable and inefficient in energy consumption. In this paper, we propose an approach where edge intelligence is imposed in a clustered and cohesive manner. This enables the IoT devices to form one or more clusters based on the intelligence they possess and operate collaboratively with or without any intervention of the remote control center. This allows IoT devices to efficiently collect and react to the environment, resulting in better service quality and further reducing overall energy consumption.","PeriodicalId":274131,"journal":{"name":"2022 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121039444","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：