Collaborative Research: CNS Core: Small: Accelerating Serverless Cloud Network Performance

无服务器计算使云编程发生了革命性的变化，并有望成为下一个占主导地位的云计算范例。无服务器计算的主要吸引力之一是“灵活的自动伸缩”，它既能带来效率优势，又能带来经济优势。然而，这种灵活性伴随着新的网络性能挑战，这些挑战是无服务器计算环境独有的。例如：(1)内部功能链通信成为无服务器应用的瓶颈，以及(2)统一功能网关增加了功能链中的间接连接数量，从而进一步损害了网络性能。该项目提出了一项跨层次的工作，旨在通过探索多种协同策略来减少功能链通信中的延迟，从而优化无服务器云环境中的功能链通信。建议的解决方案包括(1)新的基于Quic的网络底层，无需修改租户代码即可同时提高性能和安全性；(2)使用数据平面编程和知情请求预测来优化资源分配并缓解无服务器多租户云中的冷启动问题。这项工作将支持一大类具有严格延迟和安全要求的新型无服务器应用程序，我们建议的设计将普遍适用于流行的云平台。该项目将使提议的实现开源，并通过NSF交换矩阵试验台公开发布我们的评估环境，从而使不同的利益相关者受益。该项目将积极探索我们的工作向商业云服务和开源无服务器社区的过渡，如OpenFaaS和Quic-Go项目。该项目将包含一个详细的教育和推广计划，面向本科生和研究生研究人员(通过课程和暑期实习)，以及妇女和少数群体(通过ThriveWiSE等组织)。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Serverless computing has revolutionized cloud programming and is poised to become the next dominant cloud computing paradigm. Among the major allures of serverless computing is “agile autoscaling”, which brings forth both efficiency and economic advantages. However, such flexibility is accompanied by new network performance challenges that are uniquely endemic to the serverless compute environments. For example: (1) internal function chain communication becomes a bottleneck for serverless applications and (2) a unified function gateway increases the number of indirect connections in function chains that further impairs network performance. This project proposes a cross-layered effort that seeks to optimize the function chain communication in severless-cloud environments by exploring multiple, synergistic strategies to reduce latency in function-chain communication. The proposed solution includes (1) a new QUIC-based network substrate that can simultaneously improve performance and security, without the need for tenant code modification and (2) use of data-plane programming and informed request prediction to optimize resource allocation and mitigate cold-start issues in serverless multi-tenant clouds. This work will empower a broad class of novel serverless applications with stringent latency and security requirements and our proposed designs will be universally applicable across popular cloud platforms. This project will make the proposed implementations open-source and publicly release our evaluation environments through the NSF Fabric testbed, thus benefiting diverse stakeholders. This project will actively explore the transition of our work to commercial cloud services and open-source serverless communities, such as OpenFaaS and the QUIC-go project. This project will incorporate a detailed education and outreach plan targeting undergraduate and graduate researchers (through courses and summer internships), and women and minorities (through organizations like ThriveWiSE).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.