'Cloud computing' is one of the promising approaches to reduce power consumption of IT equipment. In cloud computing, IT services and the equipment to provide the services are available on the network, instead of being owned by individual users or companies. For example, you could call web mail services, such as Gmail, cloud services. These services are centrally controlled and hosted by huge data centers. The overall power consumption of IT equipment is reduced through shared use by a large number of users as well as through improved efficiency by centralization.

In achieving cloud computing, data storage devices such as disks are of particular importance. Unprecedented volumes of data must be stored and provided to enormous number of users in cloud computing. Needless to say, reliability is absolutely essential. These requirements are met by placing multiple storage devices in parallel and automatically replicating data among them.

We believe that performance guarantees are also important for the advancement of cloud computing. Guaranteeing performance is not general practice in the Internet world. However, for applications such as video streaming, which is expected to become richer and richer in the coming years, it is necessary to guarantee the performance of data storage systems, as well as networks for data transfer to the end users.

Another important topic is the semiconductor storage devices, which are rapidly becoming commoditiesies. Semiconductor storage devices provide faster access speeds and consume less power than HDDs. However, if HDDs are replaced by semiconductor storage devices and the network speed remains the same, the network speed will become a bottleneck, preventing full use of the high speed of semiconductor storage devices. We are conducting research on technology to take full advantage of semiconductor storage devices by minimizing data transfer.

With the increasing number of information systems being operated in data centers in recent years, power consumption and heat generation at data centers have become a major concern. In particular, the server density per unit area is increasing, resulting in increased power consumption. There are, however, seasonal, weekly, and daily changes in the operation of most servers used for business applications. Their capacity is not fully utilized, with an average utilization rate of only about 30 %. From the perspective of energy saving, it is more efficient to turn off the power of servers not in use and increase the utilization of those that are turned on as much as possible.

With the advent and dissemination of Software as a Service (SaaS) and cloud computing, users can enjoy a variety of services without being aware of the physical servers. In data centers providing these services, physical servers can be logically divided into several virtual servers to increase their utilization, employing virtualization technology. In this way, the number of physical servers used to provide the same service can be reduced, contributing to reductions in energy consumption. In reality, however, there is a limit to the improvement of energy efficiency using the current level of virtualization technology. The utilization of data centers generally varies with time, making it difficult to constantly maintain high utilization.

We are developing a cross-data-center virtualization technology as an advanced solution to improve energy efficiency. This technology allows data centers remotely located from each other to be managed in a unified manner through an interconnecting network as if they were a single data center. It also allows a service being operated by a data center at one location to be moved to a data center at another location without interrupting the service.

Within a data center, a technology called live migration of virtual machines is used to change the operating location of services. However, it was believed to be difficult to use this technology between remotely located data centers due to performance degradation resulting from delays in the network. We have solved this problem by proposing a new data transfer technology to overcome such delays. Performance degradation due to a delay in the network is prevented by transparently moving storage data on a virtual computer to a remote location. A service can be moved in less than one second by optimizing the method of moving memory data on the virtual computer.