Researchers use labels to keep cloud computing systems in order and under control

If you rely heavily on your computer or mobile devices day in and day out, you are probably using some form of cloud computing, whether it is a streaming media like Netflix, an online application like Google Gmail, or a cloud file storage like Dropbox. As a novel kind of service delivery emerged only a decade ago, cloud computing has saved considerable costs for service providers, small companies in particular, who have been struggling for better resource efficiency and quality-of-service (QoS) along the way.

To tackle the challenges within cloud computing, researchers have proposed a new solution, which was published on September 1 in Intelligent Computing, a Science Partner Journal.

It is called labeled von Neumann architecture (LvNA), an alternative computer architecture for the low-entropy cloud (LEC) stack. LEC was first brought up by Zhiwei Xu and Chudian Li of the Chinese Academy of Sciences (CAS) as the next generation of cloud computing. It is a technology stack that works from the bottom up -- from the underlying hardware to the higher-up software, letting applications within a cloud system perform uninterrupted, efficiently, and in terms of resource sharing, orderly. The term entropy, borrowed from physics, measures the disorder of a system. The higher the entropy, the greater the disorder.

“The key idea of labeled architecture is inspired by software-defined networking [SDN] in which the control plane is decoupled from the data plane to allow the control plane to be made programmable,” said the CAS researchers Chuanqi Zhang, Sa Wang, and Zihao Yu, together with other co-authors. “We find that a computer can be viewed as a network, and we can apply the principle of SDN to computer architecture.”

With a new programming interface, LvNA incorporates a set of label-powered control mechanisms. Specifically, with the help of labels, LvNA enables shared components to differentiate, isolate, and prioritize (DIP) user-defined application requests when competing for hardware resources. It attempts to identify and eliminate the “disorderly sharing” in hardware and guarantee the performance of applications.

“The root cause of disorderly sharing is that the low-level hardware lacks software semantic information,” the researchers pointed out. “Therefore, we consider that a method should be developed to convey the software semantic information down to hardware, in order to mitigate resource contentions at the hardware level.”

Labels are the core concept of LvNA. They contain at least two kinds of information, i.e., identification (which application the request belongs to) and policy (how to handle the request). LvNA attaches such a semantic label to every data access request, and the former is carried by the latter to their destination. “Labels determine when and how much resources certain requests can access and occupy, rather than all the requests competing in a disorderly fashion for all kinds of resources,” explained the researchers.

To demonstrate the effectiveness of LvNA, the researchers tapeouted Beihai, a 1.2 GHz 8-core RISC-V processor based on the LvNA architecture. Experimental results showed that Beihai was able to realize fully hardware-supported virtualization and drastically reduce performance degradation: it reduced the performance degradation caused by memory bandwidth contention from 82.8% to 0.4%, for instance.

As a general computer architecture for LEC, LvNA can be implemented in many different ways. “In future work, LvNA could be extended to carry more information across software and hardware by labels,” according to the researchers.

Chuanqi Zhang, Sa Wang, and Zihao Yu contributed equally to this work.

This work is supported partially by the National Key R&D Program of China (2016YFB1000201), the National Natural Science Foundation of China (Grant No. 62090020 and 62172388), Youth Innovation Promotion Association of the Chinese Academy of Sciences (Grant No. 2013073 and 2020105), and the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDC05030200).