Power-saving server systems (PSSs) conserve power by using dedicated management servers to perform load monitoring and
measurement processing for the virtual servers that provide services to clients, and by controlling both real and virtual servers to optimize
server system configurations. However, because such systems must be capable of recovering successfully if a management server failure
occurs, PSS control methods tend to become complicated, and management loads tend to increase according to the number of targeted
servers. With those points in mind, the present study proposes a PSS that does not require a dedicated management server. Herein, the
characteristics of the conventional and proposed methods used in PSSs are described and compared. To solve the problems caused by
adopting the proposed method, executor and synchronization systems are introduced, and details of their operation methods are presented.
In addition, power-saving states are set for two virtual server configurations (redundant and non-redundant) that can be selected based on
actual server system operating states. Detailed operations governing load measurement and configuration changes in the PSS are also
shown, and it is demonstrated that processing will continue even if the targeted virtual server fails during operation. Furthermore, an
experimental system using the proposed method is constructed, and its architecture, specifications, and detailed operation procedures are
described. In the power-saving condition, the proposed system operates at approximately 72% of the power consumption seen under
normal conditions when the virtual servers are configured redundantly, and at approximately 40% of normal power consumption when a
non-redundant configuration is used. Therefore, it is expected that the operating costs and operation loads of the management and target
servers can be reduced by the power saving effects and failure avoidance countermeasures of our proposed system.
View full abstract
