Extensible Operating Systems

• Aegis/Exo-kernel (Massachusetts Institute of Technology)
  Group Members: Frans Kaashoek, Dawson Engler
  The Aegis kernel is built around the idea of an exo-kernel. An evolution from micro-kernels, exo-kernels export a virtual machine that securely multiplexes resourses among mutually distrusting spplications. The exo-kernel philosophy tries to export as few abstractions besides the basic hardware abstractions as possible, and to implement as little policy in the kernel as possible. Like SPIN, Aegis relies on techniques such as downloading code into the kernel to make the system fast.
• Cache kernel (Stanford University Distributed Systems Group)
  Group Members: David Cheriton, Kenneth Duda
  The supervisor-mode component of the V++ operating system. The Cache Kernel caches operating system objects such as threads and address spaces just as conventional hardware caches memory data. User-mode application kernels handle the loading and write-back of these objects, implementing application-specific management policies and mechanisms.
• Choices (University of Illinois, Urbana-Champaign)
  Choices is written as an object-oriented operating system in C++. As an object-oriented operating system, its architecture is organized into frameworks of objects that are hierarchically classified by function and performance. The operating system is customized by replacing subframeworks and objects. The application interface is a collection of kernel objects exported through the application/kernel protection layer. Kernel and application objects are examined through application browsers
• Flux (University of Utah) Group Members
  The Flux Project's objective is to develop a nanokernel-based decomposed operating system that achieves high performance while retaining inter-component protection and rich functionality. Such a system will overcome the performance/protection and performance/functionality tradeoffs that thwart traditional micro-kernel-based operating systems. This objective includes integration of selected research results of others, and free distribution of an unencumbered and usable version of the entire system.
• Kea (University of British Columbia)
  Group Members: Norm Hutchinson, Alistair Veitch, Peter Smith, et al.
  Kea is an operating system architecture designed to provide the applications running on top of it with the ability to dynamically replace parts of the system. Don't like your file-systems buffering policy? Write your own, the system will replace it for your application! Of course, there are lots of research and design issues over how we do this on a fine-grained scale, whilst maintaining system integrity and efficiency.
• Protection and Sharing through Shared Libraries (University of Notre Dame)
  Group Members: Arindam Banerji
  Protected Shared Libraries (PSLs) are based on the observation that shared libraries, with some modifications, could be used to implement user-level OS services with adequate protection and many degrees of sharing. Furthermore the dynamic nature of loadable shared libraries can be extended to provide a high degree of flexibility in the implementation of user-level services. Protected Shared Libraries add two different dimensions to the usual notion of shared libraries - protection domains and multiple forms of sharing through Shared Address Libraries.
• Scout (University of Arizona) Group Members
  The Scout project is a new operating system project that aims to build a fast, customizable operating system for networked systems by looking at novel ways to structure and construct operating systems. The Scout operating system is designed around the path, which is how data flows between end-points in a system. Paths are primary objects to which resources are allocated in Scout. The Scout system is also exploring new compiler technologies for system design and implementation, based on the insight that extensible operating systems are worthless if no one can build or extend them.
• SPACE (University of California, Santa Barbara)
  Group Members: John Bruno, Dave Probert
  SPACE is an approach to operating systems which uses multiple protection domains rather than a single kernel to provide operating system services. Multiple instances of fundamental paradigms, such as threads and virtual memory, can coexist, since they are implemented as applications code. All that is left in what was the operating system kernel is a set of mechanisms to implement the protection domains. In SPACE these mechanisms.can be replaced as needed by the application to provide a fundamental level of extensibility not available in other adaptive operating systems.
• SPIN (University of Washington) Group Members
  SPIN is one of several research systems that aims toward run-time flexibility and specialization using techniques like type-safe languages and dynamic code generation to make a fast, dynamic, flexible system. It is an extensible operating system micro-kernel that supports the dynamic adaptation of system interfaces and implementations through direct application control, while still maintaining system integrity and inter-application isolation.
• Spring System (Sun)
  Sun's new research kernel. Spring is a highly modular, object-oriented operating system, which is focused around a uniform interface definition language. Spring is intrinsically distributed, with all system interfaces being accessible both locally and remotely.
• Synthesis (Columbia University)
  Group Members: Henry Massalin
  The Synthesis kernel was one of the first modern operating system projects to use run-time code generation (which Massalin called code synthesis) and fine-grained scheduling to construct a system that responded quickly and dynamically to high-speed devices. The main "draw-back" of Synthesis was that it was written is 68000 macro-assembler. Synthesis has influnced most of the modern extensible research operating systems, including SPIN, Aegis, Scout, and Synthetix.
• Synthetix (Oregon Graduate Institute)
  Group Members: Andrew P. Black, Charles Consel, Calton Pu, et al.
  The Synthetix project is investigating the application of a technique called incremental specialization, a combination of fine-grain modularity and dynamic code generation, to create operating systems which are both highly modular and high-performance. Incremental specialization takes advantage of particular circumstances, not just at compile time, but also at load time and run time, to make specialized optimizations.
• Vino (Harvard University)
  VINO's key attributes are reusability of kernel mechanisms by user-level applications, application-directed kernel policy, and an access-method paradigm similar to that used by database management systems. The goal of VINO is to reduce the size of conventional operating systems by leaving the definition of complex kernel policies with applications. Additionally, the VINO system will provide improved performance for applications, since they can retain more control over their resources. Finally, VINO is designed to enable the development of new applications, such as real-time video, that require explicit control of system resources.
• x-kernel (University of Arizona)
  The x-kernel is an object-based framework for implementing network protocols. It defines an interface that protocols use to invoke operations on one another (i.e., to send a message to and receive a message from an adjacent protocol) and a collection of libraries for manipulating messages, participant addresses, events, associative memory tables (maps), threads, and so on.