Improving the efficiency of computer center operations is an important issue. While the use of computer centers is increasing, the convenience of computer centers is decreasing as the waiting time for submitted jobs increases. Furthermore, the evaluation metrics for system optimization are changing and the complexity of the system is increasing, such as operational optimization from the perspective of power constraints and power cost containment, and operational optimization through collaboration with other computer centers and cloud computing. It is required to respond flexibly and quickly to these changes in operations and the system itself, and to achieve efficient operations.
However, the operation of the current computer center system still relies heavily on the knowledge of the administrator. There are many dashboards that inform the administrator of the system status and parameters that can change the operation of the system. However, the optimal parameter settings in line with certain operational goals still rely on the knowledge of the administrator. As a result, operating an increasingly complex and changing system is extremely difficult, and efficient operation has not been achieved.
On the other hand, recent advances in AI technology have made it possible to optimize and automate the use of data in a variety of areas. In the operation of computer centers, a large amount and variety of data continues to be generated, such as logs generated by devices that make up the system and information on submitted jobs. Expectations are growing for the automation of system operation optimization by utilizing these data.
This research group will construct a framework for problem solving by utilizing the data generated in the operation of computer centers and demonstrate problem solving by this framework, with the aim of realizing efficient operation by responding flexibly and quickly to increasingly complex and changing systems in computer centers.