CNS Core: Small: Exploring the Benefits on Non-Work-Conserving Networking

CNS 核心：小型：探索非工作保护网络的好处

• 批准号: 2006530
• 负责人: Roch Guerin
• 金额: $ 45.86万
• 依托单位: Washington University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-01-01 至 2024-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2006530&HistoricalAwards=false
• 关键词: CNS; Core; Small; Exploring; Benefits

Metering gates that pace traffic at highway entrances during peak hours have become common practice. They prevent multiple cars from entering back-to-back, which can create bubbles of congestion that propagate and generate broad disruptions. A similar principle applies in the data networks that make up the modern communication infrastructure. This project explores the use of such metering/pacing options to improve the performance of those networks. In particular, can latency-sensitive traffic such as interactive transactions, games, audio and video, etc., be paced, i.e., slowed down on entry, while improving the network's ability to deliver it in time? Answering this question is the goal of this project.Most end-points feeding the Internet are today located in datacenters overseen by large-scale providers. This offers the opportunity for better control of the traffic they generate. This project aims to develop a principled approach for selecting how to best pace end-point traffic as a function of traffic profiles and sensitivity to delay towards minimizing the network resources (and therefore cost) required. This can be formulated as an optimization problem whose complexity, however, grows rapidly with network size. The project targets practical guidelines to realize near-optimal pacing together with software solutions to orchestrate the resulting controls, and an evaluation of the benefits they afford.The recent COVID-19 crisis and the shift to working remotely and to online learning has triggered an unprecedented growth in the volume of "real-time" traffic. This has in turn stressed the network infrastructure responsible for carrying it, e.g., forcing some providers to limit the amount of traffic they allow. The approaches this project seeks to develop have the potential to allow those same networks to carry substantially more real-time traffic while preserving performance and without requiring costly upgrades. The work will be carried out in collaboration with a large-scale cloud provider to ensure it captures the practical constraints that networks connecting cloud sites face.The project's website is in the form of a wiki hosted by the Open Science Foundation and accessible at https://osf.io/mh4eg/. The wiki serves as an entry point for results produced by the project. It includes a section devoted to publications produced under the project's auspices and setup to ensure open-access to those works. Other sections will provide either repositories or links to external repositories for data sets and software produced as part of the project. Repositories will be selected based on their accessibility and ability to ensure long-term access to hosted information. Selected repositories will ensure preservation of the data for at least five years beyond the project's end, and preferably longer.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.

计量大门，高峰时段的高速公路入口步伐已成为普遍做法。 它们阻止了多辆车进入背靠背，这可以产生拥塞的气泡，从而传播并造成广泛的破坏。 类似的原则适用于组成现代通信基础架构的数据网络。 该项目探讨了使用此类计量/起搏选项来改善这些网络的性能。 特别是，可以节奏延迟敏感的流量，例如交互式交易，游戏，音频和视频等。回答这个问题的目标是该项目的目标。喂食互联网的大多数终点位于大型提供商监督的数据中心中。 这为更好地控制其产生的流量提供了机会。 该项目旨在开发一种原则性方法，用于选择如何最佳速度端点流量作为流量配置文件的函数和敏感性，以延迟最大程度地减少所需的网络资源（因此成本）。 这可以作为一个优化问题进行表述，但是，随着网络大小的速度，复杂性迅速增长。 该项目针对实用的准则，以实现近乎最佳的节奏以及软件解决方案来协调所得控制，并对他们所获得的收益进行评估。最近的Covid-19危机以及远程工作和在线学习的转变触发了“实时”流量的前所未有的增长。 反过来，这强调了负责携带它的网络基础架构，例如迫使一些提供商限制其允许的流量。 该项目试图开发的方法有可能使这些网络在保持性能的同时提供更多的实时流量，而无需昂贵的升级。 这项工作将与大型云提供商合作进行，以确保它捕获连接云站点面对的网络的实际约束。该项目的网站是由开放科学基金会托管的Wiki的形式，可在https://osf.io/mh4eg/上访问。 Wiki是该项目产生的结果的入口点。 它包括一个专门针对项目主持下制作的出版物的部分，以确保对这些作品开放访问。 其他部分将提供存储库或链接到外部存储库，以作为项目一部分生产的数据集和软件。存储库将根据其可访问性和确保长期访问托管信息的能力选择。 选定的存储库将确保在项目结束之前至少五年保存数据，并且最好使用更长的时间。该奖项反映了NSF的法定任务，并被认为是值得通过基金会的知识分子优点和更广泛的影响来通过评估来支持的。

项目成果

Adaptive Edge Offloading for Image Classification Under Rate Limit

• DOI: 10.1109/tcad.2022.3197533
• 发表时间: 2022-07
• 期刊: IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
• 影响因子: 2.9
• 作者: Jiaming Qiu-;Ruiqi Wang;Ayan Chakrabarti;R. Guérin;Chenyang Lu

Minimizing network bandwidth under latency constraints: The single node case

在延迟限制下最小化网络带宽：单节点情况

• 发表时间: 2021
• 期刊: Proceedings 33rd International Teletraffic Congress (ITC 33
• 作者: Jiayi Song, Roch Guérin