CNS Core: Small: Designing Networks for Stringent Performance Requirements

随着在线和基于云的服务的广泛采用，底层计算机网络基础设施满足日益严格的性能要求（例如，维持业务关键型应用程序吞吐量）。这些要求必须得到满足，尽管故障是互联网服务提供商（ISP）和云提供商网络的全球规模及其快速发展的常态。该项目将通过开发新技术来应对这些挑战，这些新技术使网络架构师能够规划其网络设计（例如，设计拓扑结构、提供容量、故障时重新路由流量的算法），以便在网络可能遇到的大量场景中性能都是可接受的。该项目将汇集网络系统和优化理论方面的专业知识，并在两个关键方面推进最新技术。首先，该项目将开发（i）可在网络中部署以响应故障的新机制，以及（ii）可以证明所产生的性能在指定故障状态下是可接受的框架。该框架的一个新奇是能够对丰富而灵活的网络机制进行建模。第二，与现有的方法只考虑最坏情况下的性能，可能过于保守，这个项目将开发新的方法来设计的要求，必须满足所需的百分比的情况下，该项目将使重大的现实世界的影响，通过确保网络符合服务水平目标，在面对故障，并导致网络具有更低的成本，更好的性能，和更高的可靠性。通过在优化方面取得根本性进展，该项目将影响网络之外的其他科学学科。该项目将参与行业和网络运营商论坛，并广泛参与博士，硕士和本科生的研究。该项目将在网络和优化课程中纳入跨学科材料，并将积极参与K-12外联活动。https://engineering.purdue.edu/~isl/certifiabledesign.html该奖项反映了NSF的法定使命，并被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

With the wide-spread adoption of online and cloud-based services, it is critical that the underlying computer network infrastructure meet increasingly stringent performance requirements (e.g., sustain throughput for business-critical applications). These requirements must be met despite failures that are the norm given the global scale of Internet service provider (ISP) and cloud provider networks, and their rapid pace of evolution. This project will tackle these challenges by developing novel techniques that enable network architects to plan their network designs (e.g., design topology, provision capacity, algorithms for re-routing traffic on failures) so that performance is acceptable over a large set of scenarios the network may encounter.The project will bring together expertise in network systems, and optimization theory, and advance the state-of-the-art in two key ways. First, the project will develop (i) new mechanisms that may be deployed in the network to respond to failures, and (ii) frameworks that can certify that the resulting performance is acceptable over specified failure states. A novelty of the framework is the ability to model rich and flexible network mechanisms. Second, unlike existing methods that only consider worst-case performance, and may be overly conservative, this project will develop novel ways to design for requirements that must be satisfactory for a desired percentage of scenarios.The project will enable significant real-world impact by ensuring that networks comply with Service Level Objectives in the face of failures, and lead to networks with lower cost, better performance, and higher reliability. By making fundamental advances in optimization, the project will impact other scientific disciplines beyond networking. The project will engage with industry and network operator forums, and extensively involve doctoral, Masters, and undergraduate students in the research. The project will incorporate inter-disciplinary material in networking and optimization courses, and will involve active participation in K-12 outreach activities.The results from the project will be available at: https://engineering.purdue.edu/~isl/certifiabledesign.html. The results will be available throughout the duration of the project, and for three years after.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.