II-NEW: One Cloud Does Not Fit All: Minnesota Integrated Cloud Systems Research Testbed (MiST)

在过去的几年中，大型数据中心的大量计算资源和存储容量出现了巨大的增长，使新的互联网应用程序和服务成为可能。与此同时，智能手机、平板电脑和电子阅读器也得到了快速增长和广泛采用，用户可以在移动中访问这些应用程序和服务。这些互补趋势的融合产生了移动服务和云计算等新兴概念。然而，当前的移动服务和云计算产品为实现其全部潜力带来了许多困难的挑战。这些挑战包括缺乏用户对计算资源的控制，由于用户设备和云资源之间的分离而导致的性能差和缺乏上下文，以及集中式云提供的有限隐私。这些挑战提出了一些重要的问题，例如：我们能否建立大规模的、以社区为基础的建筑？通过利用大量小型企业提供的计算和存储资源，实现可定制的云计算基础设施？我们能否在利用现有云计算基础设施的同时，开发个性化和协作的互联网云服务和移动应用程序，使用户能够控制？哪些应用程序可以在每种类型的云上有效运行，这些云如何耦合以产生更强大的平台？我们如何才能充分发挥移动和云计算的潜力，同时最大限度地减少其碳足迹？明尼苏达大学（University of Minnesota， UMN）的pi正在开展广泛的研究，以解决这些挑战，通过构建一个多样化和分散在广泛地区的集成云系统；它不仅利用商业大规模云基础设施，还利用边缘的小规模本地社区资源。这些研究工作继续取得进展的一个重大障碍是缺乏用于实验的灵活可控的计算、存储和网络基础设施。该项目旨在通过建立明尼苏达集成云系统测试平台（MiST）研究基础设施来推进云计算基础设施和应用程序的研究，从而克服这一障碍。该研究设施分为三个相互关联的逻辑层：i)由两个数据中心组成的核心云系统；Ii)由几台服务器机器组成的多个分布式边缘服务器云；以及iii)固定和移动用户群，包括各种不同的用户设备。在教育方面，这个研究基础设施提供了一个理想的“动手”学习环境，教授研究生和本科生重要的数据分析、系统构建和实验技能，这些技能对当今的IT劳动力至关重要。pi将MiST研究基础设施的研究纳入课堂教学，包括网络、安全、存储和分布式系统等核心课程的课堂项目，以及本科生、高年级和定向研究课程。私立学院在支持本科生方面有着良好的记录，尤其是女性和代表性不足的学生。pi计划通过与其工业合作者的密切互动，以及通过出版物、演讲和向更大的研究社区公开发布研究数据、软件工具和原型系统来传播研究进展。

The past few years have seen a tremendous growth of massive computing resources and storage capacities in large data centers enabling new Internet applications and services. At the same time, there has been a rapid growth and widespread adoption of smart phones, tablets, and e-readers, through which users access these applications and services on the move. The confluence of these complementary trends has given rise to the emerging notions of mobile services and cloud computing. However, current mobile services and cloud computing offerings pose many difficult challenges to realize their full potential. These challenges include lack of user control over computing resources, poor performance and lack of context due to the separation between user devices and cloud resources, and limited privacy provided by centralized clouds. These challenges raise important questions such as: Can we build large-scale, ?community-based?, customizable cloud computing infrastructures by leveraging computing and storage resources contributed by a large number of small players? Can we develop personalized and collaborative Internet cloud services and mobile apps that enable user control, while tapping into existing cloud computing infrastructures? What applications can be run effectively on each type of cloud, and how could such clouds be coupled to yield a much more powerful platform? How can we fully realize the potential of mobile and cloud computing while minimizing its carbon footprint?The PIs at the University of Minnesota (UMN) are conducting a wide range of research on addressing these challenges, by building an integrated cloud system that is diverse and dispersed across a wide area of locations; it not only utilizes the commercial large-scale cloud infrastructures, but also leverages smaller-scale, local community resources at the edge. A significant barrier to continued progress in these research efforts is lack of a flexible and controllable computing, storage and networking infrastructure for experimentation. This project seeks to overcome this obstacle by building the Minnesota Integrated cloud Systems Testbed (MiST) research infrastructure for advancing research on cloud computing infrastructures and applications. This research facility is organized in three interconnected logical tiers: i) core cloud systems consisting of two data centers; ii) several distributed edge server clouds consisting of a few server machines; and iii) swarms of stationary and mobile users, comprising a wide array of different user devices. In terms of education, this research infrastructure provides an ideal "hands-on" learning environment to teach both graduate and undergraduate students important data analysis, system building and experimental skills that are critical for today's IT workforce. The PIs incorporate research conducted on the MiST research infrastructure in their classroom teaching, both for class projects in the core courses on networking, security, storage, and distributed systems, as well as in undergraduate senior and directed research courses. The PIs have a strong track-record of supporting undergraduate students, especially women and under-represented students. The PIs plan to disseminate the research advances via close interaction with their industrial collaborators, and through publications, presentations and public release of research data, software tools and prototype systems to the larger research community.