数据操作系统的思想、设计和原型验证
摘要
The modern operating system (OS) is inclined to be less dependent on devices and more involved with data. Data manipulation now tends to be the core functionality of future OSes. This is also one principle of the sea-cloud computing environment. By analyzing the challenges and problems of traditional OSes, this paper proposes the design and implementation paradigm of a next-generation OS, DataOS. DataOS introduces data abstraction and an in-kernel data stack that can directly support data collection, transfer, processing, storing, and service-providing. DataOS presents a novel solution not only for performance improvement of current big data systems but also for operating system evolvement. As a proof of concept, we implement the first version of the DataOS prototype that includes a data stack and ports the Key/Value storage system. The results of experiments confirm that DataOS outperforms the classical Memcached in terms of both real-time and concurrency and validate the feasibility and efficiency of the philosophy and design of DataOS.