VELOSITY: A New Architecture for Operating System Design and Implementation

影响我们的计算基础设施是建立在一个50年的软件架构。Velocity计划是系统设计和实现的一种大胆方法，利用了过去50年在操作系统架构、程序合成和安全性方面的进步。由于超大规模集成电路（VLSI）实现了前所未有的硬件创新和发展，Velocity满足了当今对灵活、可靠和安全的系统软件的需求，释放了未来硬件平台的力量。在过去的三十年里，我们的硬件基础设施一直基于x86单一文化。根据图灵奖得主大卫帕特森的说法，摩尔定律的终结产生了“计算机架构的一个新的黄金时代。“这个新时代对系统软件提出了新的要求：跨硬件实现的可移植性和支持新兴平台的能力。Velocity的长期目标是开发操作系统组件库，这些组件来源于机器独立的规范，包括功能，安全和内省功能。从这些正式指定的安全组件中，我们可以构建定制的操作系统或系统软件。因此，具有新硬件的供应商可以组装具有精确所需功能的系统软件。Velocity通过结合领域特定语言（DSL）和程序合成来解决当今的硬件异构性问题。操作系统设计者在专门为此目的设计的DSL中指定小组件所需的功能和安全属性;硬件供应商使用不同的DSL描述其硬件或硬件接口的语义。通过将这两个DSL与最先进的程序合成和编译技术相结合，我们可以为各种硬件平台生成系统软件，而无需在平台之间移植（翻译）低级系统的繁琐过程。Velosity需要研究操作系统架构、DSL开发和程序合成。在这里，我们解决了四个挑战：1）设计和开发一个组件库，我们可以构建系统软件。2)用于指定系统功能的DSL的设计和实现，以及3）硬件能力。4)对现有的程序综合技术进行调整，并为系统软件开发新技术。在为期五年的探索基金期间，我将培养两名博士生，五名硕士生和一名博士后研究员。总而言之，超大规模集成电路技术给硬件设计带来了两个基本的好处，使可重用组件库的开发和体系结构实验成为可能。Velocity为系统软件提供了相同的基本优势。它的组件库是可重用的组件，这将允许在异构硬件上进行架构实验。因此，Velocity为硬件平台提供了正确的系统软件堆栈。

Impact Our computing infrastructure is built on a fifty year old software architecture. The Velosity program is a bold approach to system design and implementation, leveraging the past fifty years of progress in operating system architecture, program synthesis, and security. As Very Large Scale Integration (VLSI) enabled unprecedented hardware innovation and evolution, Velosity addresses today's need for flexible, reliable, and secure system software, unleashing the power of tomorrow's hardware platforms. For the past thirty years, our hardware infrastructure has been based upon the x86 monoculture. According to Turing laureate David Patterson, the end of Moore's Law produced "A new golden age for computer architecture." This new era brings new demands for system software: portability across hardware implementations and the ability to support emerging platforms. Objectives, Methodology and HQP The long-term objective of Velosity is to develop libraries of OS components, derived from machine independent specifications, that include functional, security, and introspective capabilities. From these formally specified, secure components we can construct customized operating systems or system software. Thus, a vendor with new hardware can assemble system software with precisely the needed functionality. Velosity addresses today's hardware heterogeneity through a combination of domain specific languages (DSL) and program synthesis. Operating system designers specify the required functionality and security properties of small components in a DSL designed specifically for that purpose; hardware vendors describe the semantics of their hardware or hardware interfaces using a different DSL. By combining these two DSLs with state-of-the-art program synthesis and compilation techniques, we produce system software for a wide range of hardware platforms without the tedious process of porting (translating) low-level systems between platforms. Velosity requires research in operating system architecture, DSL development, and program synthesis. Here, we address four challenges: 1) Design and development of a component library from which we can construct system software. 2) Design and implementation of DSLs for specifying system functionality, and 3) hardware capabilities. 4) Adaptation of existing program synthesis techniques and development of new techniques for system software. Over the five-year Discovery Grant term, I will train two PhD students, five MSc students, and one postdoctoral fellow. To summarize, VLSI technology brought two fundamental benefits to hardware design, enabling both development of libraries of reusable component and architectural experimentation. Velosity provides the same essential benefits to system software. Its component libraries are reusable components, which will allow for architectural experimentation on heterogeneous hardware. Thus, Velosity provides hardware platforms with the right system software stack.