CAREER:Exploiting Gray-Box Techniques in Systems

现代操作系统是大型而复杂的代码库，在这方面已经投入了数百年的编程时间。因此，修改操作系统是一项困难、昂贵且往往不切实际的工作。然而，将操作系统视为不变的对象与大多数操作系统研究不一致。这项研究提出了能够在不改变操作系统源代码的情况下部署操作系统的技术。论文的主题是，类似操作系统的服务可以获取关于操作系统内部状态的信息并控制其行为，即使没有提供显式的接口来这样做。通过这种方法，操作系统被视为一个灰盒，在其中其算法的一般特征是已知的；然后，该知识与操作系统如何反应的运行时观察相结合，以推断其状态并施加控制。灰盒系统的初步经验指向其他研究领域：算法的自动发现、配置参数的在线基准测试、动态插入探测器以及积极推断操作系统的内部状态。因此，这项提议的目标是双重的：了解灰箱系统的理论基础，并为灰箱系统的快速开发实现一个软件工具箱。

Modern operating systems are large and complex code bases in whichhundreds of programmer-years have been invested. As a result,modifying an operating system is a difficult, costly, and oftenimpractical endeavor. However, viewing the OS as an immutable objectis at odds with most OS research.This research proposes techniques that enable the deployment of OSideas without changing OS source code. The thesis is that an OS-likeservice can acquire information about the internal state of the OS andcontrol its behavior, even when no explicit interfaces to do so areprovided. With this approach, the OS is treated as a gray box, inwhich the general characteristics of its algorithms are known; thisknowledge is then combined with run-time observations of how the OS reactsto probes in order to infer its state and exert control.Initial experience with gray-box systems points to additional areas ofresearch: automated discovery of algorithms, on-line benchmarking toconfigure parameters, dynamic insertion of probes, and aggressiveinference of OS internal state. Thus, the goal of this proposal istwo-fold: to understand the theoretical foundations of gray-boxsystems and to implement a toolbox of software for the rapiddevelopment of gray-box systems.