CAREER: Advancing Experimental Computer Science in Storage Management and Education

这项工作将通过使用实验技术来推进自动存储分配，并将促进实验计算机科学的教育。动态分配的存储空间的自动回收称为“垃圾回收”(GC)。GC有助于避免各种常见但严重的编程错误。不幸的是，这可能会对程序性能产生不利影响。现代收集器可能提供了可以接受的性能，但缺乏实验证据来支持这一说法，这阻碍了GC的采用，尽管它具有优势。这项工作的目标是更好地理解GC性能，既在性能足够时帮助采用它，又在性能不足时提供有关如何改进性能的详细定量信息。特别是，这项工作将比较复制和标记和清除收集在现代计算机体系结构上的性能。工作的重点将是建立GC试验台，这将有助于对各种GC技术进行系统的比较研究。其教学目标是提高计算机实验教学水平。建议开设一门新课程，明确教授实验型计算机科学。教学方法是用于教授其他实验科学的标准方法，在这种方法中，著名的实验重复进行，使学生能够获得基本方法和技术的实践经验。

This work will advance automatic storage allocation through the use of experimental techniques, and will advance education in experimental computer science. Automatic deallocation of dynamically allocated storage is called "Garbage collection" (GC). GC helps to avoid a variety of common but serious programming errors. Unfortunately, it may adversely affect program performance. Modern collectors probably provide acceptable performance, but there is a lack of experimental evidence to support this claim, hindering the adoption of GC, despite its advantages. The goal of this work is to provide a better understanding of GC performance, both aiding its adoption when performance is adequate, and providing detailed quantitative information about how performance needs to be improved when it is inadequate. In particular, the work will compare the performance of copying and mark-and-sweep collection on modern computer architectures. The focus of the work will be building a GC testbed, which will facilitate the systematic comparative study of a variety of GC techniques. The educational goal is to improve the teaching of experimental computer science. A new course is proposed that will explicitly teach experimental CS. The pedagogiaal approach is the standard one used to teach other experimental sciences, in which well-known experiments are repeated, enabling students to gain hands-on experience with the basic methodologies and techniques.***