An integrated cyber-physical system for automatic identification of massive discs
Yao Jie1, Zhang Yifan1, Cao Qiang1,2, Xie Changsheng2     
1. College of Computer Science, Huazhong University of Science and Technology, Wuhan, Hubei 430074, China;
2. Wuhan National Laboratory for Optoelectronics, Wuhan, Hubei 430074, China

Overview: The growing volumes and potential values of Big Data demand the long-term preservation of such data at extremely low cost. Optical discs are archival storage media with long lifetime and high environmental tolerance but very low $/GB cost. Furthermore, discs can store data safely and almost perpetually in an offline manner. Discs used in frequently can be simply placed offline on the common bookshelves without deploying expensive servers and power utility, and environment controlling equipments, costly provisions, which are mandatory for hard disks and tapes. Therefore, discs are very suitable for massive cold-data preservation in long-term.

Only when accessed, discs need to be loaded into optical drives to be accessed online. However, the premise of enabling an effective physical switch of discs between the online optical drives and offline locations (e.g., bookshelves) heavily depends on an automatic disc identification mechanism to accurately and efficiently locate a requested disc. As a consequence, each disc should be given a uniform and unique identifier (ID) in both cyber and physical worlds. The physical ID can be recognized by machines or human beings. The logical (cyber) ID is a traceable metadata set containing identifying information, such as a complete certifications chain. The logical ID is accessed by optical drives, hereafter referred to as burners. The automatic disc identification system needs to print physical identification on discs by printer and further to burn the logical certification and identifies certification in discs by burner. In tradition, both the burner and printer are two completely independent devices. It is needed to fetch optical discs between the printer and burner mechanically and precisely. Therefore, we are also strongly motivated to design and implement a mechanical-electrical synthesis iCPS built upon off-the-shelf printer, burner, and camera.

This paper designed a batch-disc automatic identification system, which integrates common optical disc burners, printers, and cameras, automatically to print physical label and to burn logical identification on each disc. Consider that each commodity component has its own internal independent timing control and a specific external interface. Future domain-specific iCPS can also benefit from our proposed infrastructure and relevant approach to quickly realize the targeted function at low cost and high reliability. This study designed and developed a customized mechanical structure, as well as a global software scheduling mechanism to coordinate physical behavior and logical control.

We experimentally evaluate the performance and effectiveness of ADIS in a quasi-production environment. ADIS continuously and automatically identifying a stack of 200 discs at an average time of 2 minutes per disc. The overall utilization of mechanics is almost 63%. The total cost of ADIS, including the host, printer, and burner, is less than 400$. ADIS can easily scale out for high concurrent scenarios or large-scale applications.

Supported by Creative Research Group Project of National Natural Science Foundation of China (NSFC) (61821003), Key Project of NSFC (61432007), and Surface Project of NSFC (61872156)