October 2, 2019
AT TOKYO, Inc.
Keio University
AT TOKYO, Inc. (Headquarters: Koto-ku, Tokyo; President & CEO: Akira Nakamura), the Matsutani Laboratory at the Department of Information and Computer Science, Faculty of Science and Technology, Keio University (President: Akira Haseyama), Associate Professor Masaaki Kondo of the System Information 8th Laboratory, Graduate School of Information Science and Technology, The University of Tokyo (Location: Bunkyo-ku, Tokyo; President: Makoto Gonokami), and SECOM CO., LTD. (Headquarters: Shibuya-ku, Tokyo; President: Ichiro Ozeki) have launched a proof-of-concept at an AT TOKYO data center. The project aims to establish technology for anomaly detection and operational support in data center facilities using machine learning.
AT TOKYO, which handles the BPO and ICT business for the SECOM Group, supports its customers' businesses by ensuring the stable, 24/7 operation of its data centers, which are critical infrastructure supporting society.
Meanwhile, Keio University and The University of Tokyo are researching highly real-time anomaly detection technology that runs on IoT devices and does not require training data.
And SECOM is researching technology to detect medium- to long-term trend changes by comprehensively analyzing multiple time-series data, aiming for applications in various fields such as facility monitoring and surveillance.
These four parties have been working together for about a year to analyze data collected from data center facilities. As a result, it was confirmed that each organization's technology is effective to a certain degree in areas such as the early detection of anomalies that could lead to equipment failure—something difficult with conventional threshold monitoring—and the discovery of environmental changes accompanying equipment installation by data center users.
Therefore, the parties have decided to collaborate on this industry-academia proof-of-concept to establish elemental technologies for anomaly and change detection, and to aim for the practical application of these technologies for operational support, such as predictive failure detection. In this proof-of-concept, they will verify machine learning technologies applicable to services requiring high reliability, targeting data centers. In the future, they will proceed with further studies, including expanding the scope of application beyond data centers.
Please see below for the full press release.