CC-NIE Integration: Science Slices Converting Network Research Innovation into Enhanced Capability for Computational Science and Engineering at the University of Utah

随着对校园网络的需求加剧--其中许多是由科学和工程学科驱动的--传统的网络管理方法越来越过时。研究人员通常要求网络路径具有更高的吞吐量、更高的安全性和/或绕过外围防火墙。传统上，网络运营商通过提供专用资源来满足少数研究用户的需求。以前处理区域和国家设施特别连接请求的办法没有得到很好的推广。在犹他大学，一个研究团队正在使用最初为国家网络研究试验台(NSF GENI)开发的技术来升级部分校园网络。这种“可分割网络”模型允许在同一设备上并行建立多个虚拟网络。产生的每个科学切片可能具有不同的特征，并且可能连接到校园内外的不同资源。除了网络路径之外，该分配方案还可以结合计算和存储资源。由于硬件支持新兴的软件定义网络标准，这些切片可以控制到非常精细的粒度。这项技术本质上支持动态切片创建，以支持研究人员按需分配网络资源。这项工作提供了一种抽象，在校园网中干净地隔离多个高强度用户。随之而来的智力优势来自于将以研究为重点的网络管理方法更广泛地应用于大型生产环境的一部分。为了产生更广泛的影响，与S大学本科生荣誉公寓的联系使这个创新的群体能够直接与切片网络互动，也可以进行有指导的实验。核心软件库和特定于项目的增强功能在开放源码许可下可用。

As demands on campus networks - many driven by science and engineering disciplines - intensify, traditional network management approaches are increasingly obsolete. Researchers often require network paths with higher throughput, increased security, and/or perimeter firewall bypass. Network operators traditionally have accommodated a few research users by provisioning dedicated resources. Previous approaches for addressing requests for special connectivity to regional and national facilities have not scaled well. At the University of Utah, a research team is upgrading part of the campus network with technology originally developed for a national network research testbed (NSF GENI). This "sliceable network" model allows multiple virtual networks to be established in parallel over the same equipment. Each resulting science slice may have different characteristics and may connect to distinct resources either on or off campus. This allocation scheme may incorporate computational and storage resources in addition to network paths. With hardware supporting emerging Software Defined Networking standards, these slices can be controlled to very fine granularity. This technology inherently supports dynamic slice creation to enable on-demand network resource allocation for researchers. This work provides an abstraction to insulate multiple high-intensity users cleanly in a campus network. The ensuing intellectual merit follows from applying a research focused network management approach more broadly to part of a large production environment. For broader impact, a connection to the University?s undergraduate honors residence allows this innovative cohort to interact directly with the sliced network and also pursue guided experimentation. The core software base and project-specific enhancements are available under open-source licensing.