低熵云计算系统

发布时间：2018-04-03 23:31

本文选题：云计算　切入点：用户体验　出处：《中国科学:信息科学》2017年09期

【摘要】：当前的云计算系统,不论是虚拟化云还是分区云,难以同时满足用户体验和系统效率需求,产业界和学术界都开始研究下一代云计算系统以应对这个难题.本文指出,这个难题的一个重要原因是计算系统熵(无序、干扰和不确定性)居高不下,并归纳了云计算系统中存在的4类无序现象.本文提出了低熵云计算系统的学术概念,刻画了其主要特点,比较了低熵云计算系统与虚拟化云和分区云在用户体验、开发效率、运行效率、资源适配方面的区别,并讨论了低熵云的新概念和新技术:(1)不同于图灵可计算性和算法可计算性的实用可计算性概念,形式化地刻画了云计算行业的"用户体验差的功能是不存在的功能"的实践经验;(2)刻画云计算系统能够实现实用可计算性的充分必要条件,即DIP猜想;(3)支持DIP猜想,即能够区分、隔离、优先化计算任务相空间,从而降低干扰,有潜力同时满足用户体验和系统效率需求的标签化von Neumann体系结构;(4)适配深度学习负载与神经网络处理器的云计算协同设计技术.
[Abstract]:Current cloud computing systems, whether virtualized cloud or partition cloud, are difficult to meet the needs of user experience and system efficiency. Industry and academia have begun to study the next generation cloud computing system to deal with this problem.This paper points out that one of the important reasons of this problem is that the entropy of computing system (disorder, disturbance and uncertainty) remains high, and concludes four kinds of disorder phenomena in cloud computing system.This paper presents the academic concept of low entropy cloud computing system, describes its main characteristics, and compares the differences between low entropy cloud computing system and virtualized cloud and partition cloud in user experience, development efficiency, operation efficiency, resource adaptation.The new concept and new technology of low entropy cloud are discussed. The concept of practical computability is different from that of Turing computability and algorithmic computability.This paper describes formally the practical experience that "the function of poor user experience is the function that does not exist" in cloud computing industry.) it describes the sufficient and necessary condition that cloud computing system can realize practical computability, that is, DIP conjecture 3) supports DIP conjecture.That is, the ability to distinguish, isolate and prioritize the computing task phase space, thus reducing interference,The tagged von Neumann architecture, which has the potential to meet both user experience and system efficiency requirements, adapts to the collaborative design technology of cloud computing based on deep learning load and neural network processor.
【作者单位】：中国科学院计算技术研究所计算机系统结构国家重点实验室;中国科学院大学;
【基金】：国家重点研发计划(批准号:2016YFB1000200) 国家自然科学基金重点项目(批准号:61532016)资助
【分类号】：TP3

【相似文献】