CNS Core: Small: Collaborative Research: HEECMA: A Hybrid Elastic Edge-Cloud Application Management Architecture

CNS 核心：小型：协作研究：HEECMA：混合弹性边缘云应用管理架构

• 批准号: 1908574
• 负责人: Flavio Esposito
• 金额: $ 20.62万
• 依托单位: Saint Louis University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-10-01 至 2024-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1908574&HistoricalAwards=false
• 关键词: CNS; Core; Small; Collaborative; Research

Application software is becoming increasingly abundant in functionality and increasingly demanding of resources, e.g., memory and compute power. This project examines how application software, e.g., a Virtual Reality (VR) based drone control application, can be partitioned and deployed over different parts of a distributed computing infrastructure, i.e., resources are managed by a hybrid of service and cloud providers. Research questions that will be answered include: what is the best partitioning of the application that balances compute and memory demand for each application component (function) and the communication needs between these functions? Which provider should be used for each application function so that it runs quickly at the lowest cost? Moreover, how should the system adapt to changes in the availability of resources to maintain a high level of quality of experience for users?The project will develop the theoretical foundations of decomposing applications over a distributed complex cyberinfrastructure. We will study various decompositions and corresponding virtualization and resource allocation services offered by different providers with resources at the edge of the infrastructure, i.e., closer to users, and at the core, i.e., deeper into the infrastructure. The goal is to find the "best" decomposition to meet users' quality of experience while reducing the cost/price for the users. The decomposed (distributed) solution will be realized by employing corresponding feedback control algorithms and game-theoretic incentives. The project will then experiment with three classes of applications to validate these theoretical foundations: streaming Augmented/Virtual Reality (AR/VR), intrusion detection, and spatiotemporal ecological forecasting. Toward this end, a prototype will be developed and evaluated over existing open cloud infrastructures (e.g., GENI, CloudLab, Chameleon), commercial clouds, as well as private clouds (e.g., Massachusetts Open Cloud).This project will advance the state of the art in the area of application software decomposition and deployment over large-scale hybrid cloud infrastructures. The theoretical foundations and experimental validation of the work will inform the design and deployment of other foreseen and unforeseen applications. The performance and cost gains will translate to more efficient use of resources and consequently, a society that is better connected and greener. Outreach efforts will include several activities: (1) the development and delivery of hands-on tutorials that involve partitioning and running application software over a distributed cyberinfrastructure managed by multiple providers, (2) the organization of annual summer camps on cyberinfrastructure operations and management for high school students, and (3) the training and mentoring of minority and under-represented students working on research related to this project. The project will maintain a website at http://csr.bu.edu/heecma for the duration of this grant and beyond. The website will contain all the publications and results of this project in the form of code, prototype, and tutorials.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

应用软件在功能上正变得越来越丰富，并且对资源的要求也越来越高，例如，内存和计算能力。这个项目研究了应用软件，例如，基于虚拟现实（VR）的无人机控制应用，可以被划分和部署在分布式计算基础设施的不同部分上，即，资源由服务和云提供商混合管理。将回答的研究问题包括：平衡每个应用程序组件（功能）的计算和内存需求以及这些功能之间的通信需求的应用程序的最佳分区是什么？每个应用程序功能应该使用哪个提供程序，以便以最低的成本快速运行？此外，系统应该如何适应资源可用性的变化，以保持高水平的用户体验质量？该项目将开发在分布式复杂网络基础设施上分解应用程序的理论基础。我们将研究不同提供商提供的各种分解以及相应的虚拟化和资源分配服务，这些服务的资源位于基础设施的边缘，即，更贴近用户，更贴近核心，深入基础设施。目标是找到“最佳”分解，以满足用户的体验质量，同时降低用户的成本/价格。分解（分布式）的解决方案将通过采用相应的反馈控制算法和博弈论激励来实现。 然后，该项目将对三类应用进行实验，以验证这些理论基础：流增强/虚拟现实（AR/VR），入侵检测和时空生态预测。为此，将在现有的开放云基础设施上开发和评估原型（例如，GENI、CloudLab、Chameleon）、商业云以及私有云（例如，马萨诸塞州开放云）。该项目将推进应用软件分解和大规模混合云基础设施部署领域的最新技术。 这项工作的理论基础和实验验证将为其他可预见和不可预见的应用程序的设计和部署提供信息。性能和成本收益将转化为更有效地利用资源，从而实现更好地连接和更环保的社会。外联工作将包括若干活动：（1）开发和提供实践教程，涉及在由多个供应商管理的分布式网络基础设施上划分和运行应用软件，（2）为高中生组织关于网络基础设施运营和管理的年度夏令营，以及（3）对从事与本项目相关研究的少数民族和代表性不足的学生进行培训和指导。该项目将在本赠款期间及以后维持一个网站http://csr.bu.edu/heecma。该网站将以代码、原型和教程的形式包含该项目的所有出版物和成果。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

P4FL: An Architecture for Federating Learning With In-Network Processing

• DOI: 10.1109/access.2023.3318109
• 发表时间: 2023
• 期刊: IEEE Access
• 影响因子: 3.9
• 作者: Alessio Sacco;Antonino Angi;G. Marchetto;Flavio Esposito

Privacy and Efficiency of Communications in Federated Split Learning

• DOI: 10.1109/tbdata.2023.3280405
• 发表时间: 2023-01
• 期刊: IEEE Transactions on Big Data
• 影响因子: 7.2
• 作者: Zongshun Zhang;Andrea Pinto;Valeria Turina;Flavio Esposito;I. Matta

Dealing With Changes: Resilient Routing via Graph Neural Networks and Multi-Agent Deep Reinforcement Learning

• DOI: 10.1109/tnsm.2023.3287936
• 发表时间: 2023-09
• 期刊: IEEE Transactions on Network and Service Management
• 影响因子: 5.3
• 作者: Sai Shreyas Bhavanasi;L. Pappone;Flavio Esposito

A Collaborative and Distributed Learning-Based Solution to Autonomously Plan Computer Networks

• DOI: 10.1109/icin56760.2023.10073505
• 发表时间: 2023-03
• 期刊: 2023 26th Conference on Innovation in Clouds, Internet and Networks and Workshops (ICIN)
• 作者: Doriana Monaco;Alessio Sacco;Enrico Alberti;G. Marchetto;Flavio Esposito

ASAP: Adaptive and Scalable Microservice Provisioning for Edge-IoT Networks

• DOI: 10.23919/wons57325.2023.10062045
• 发表时间: 2023-01
• 期刊: 2023 18th Wireless On-Demand Network Systems and Services Conference (WONS)
• 作者: Amit Samanta;Flavio Esposito;Tri Gia Nguyen