The CPU is one of the most battery-consuming elements in a smartphone. The power consumption and user experience are in a trade-off relationship. These can be controlled by the CPU clock rate. Increasing the clock rate makes the user experience better and power consumption worse. On the contrary, the user experience declined and power consumption savings is improved by decreasing the rate. Thus, the clock rate should be increased only when large CPU resources are required. In our previous work, we proposed a concept of increasing the clock rate at an invocation of a CPU-consuming and time-consuming method. However, we did not discuss whether there are method invocations that meet these conditions. In this paper, we investigate the CPU usage trend of practical Android applications and show that many applications have method invocations suitable for the proposed dynamic CPU clock rate adjustment.

In a virtualized environment using virtual computers, a two-level cache hierarchy for accessing storage devices is constructed, which consists of a host OS cache and a guest OS cache. In such an environment, if the upper cache (guest OS cache) uses LRU as a cache replacement algorithm, it is known that negative temporal locality of reference occurs in the lower cache (host OS cache).

Fully Homomorphic Encryption (FHE) is encryption with which a user can execute calculations using its encrypted data without decrypting it. Many users hesitate to upload their private data to a public cloud computing system even though the user desire to utilize the great computing power of the public cloud computing system. FHE address this issue. A user can perform the user's calculation by the cloud computing system's processing power without providing its plain data. In the case of FHE data mining based on the Apriori algorithm, the program alternately access files. In this case, the program has to access files involving long-distance seek on a usual file layout. In a previous paper, a method for improving the I/O performance of accessing multiple large files based on striping file layout. However, this method does not support placing new intermediate files near the data files during a program without involving seeking long-distance seek between data files and newly created intermediate files. In this paper, we propose a method for improving this file layout by reserving areas for intermediate files. We then evaluate the proposed method and show that the proposed method can reduce I/O time.

Artificial intelligence, especially artificial intelligence based on deep neural networks, has improved significantly. In particular, great progress has been made in natural language processing and image recognition. In the natural language processing field, many techniques and methods, such as transformer, attention, and self-attention, have been proposed and have improved this field. Recently, BERT and RoBERTa are expected to be some of the most promising natural processing technologies. Through deep learning has achieved high accuracy in various fields such as natural language processing, it has been pointed out that it lacks interpretability and explainability for decision. For addressing this issue, providing interpretability and explainable AI have been studied. For BERT and RoBERTa, discussions on understanding what language models know about language have been studied. In this paper, we discuss what the pre-trained Japanese BERT and RoBERTa language models know. We solved masked questions for Japanese students using the deep learning artificial intelligence using these language models and investigated their knowledge and the dependence of their accuracies on the domain.

This paper provides an overview of tourist information distribution system that sends information corresponding to places and times. We completed the system successfully, although it is difficult to clearly extract information of the date, time, place, and event name from non-structured data written in natural language such as the language used on the SNS. We evaluated how many pieces of information are collected.

Docker is an emerging container-based virtualization system. It provides virtualized environment with a small overhead. To present a filesystem to each container with a small overhead, it utilizes OverlayFS, which is a union mount filesystem. In our previous work, we proposed a method for improving file access performance of OverlayFS by reducing the frequency of synchronization. In this paper, we evaluate the I/O performance in containers of Docker and show that the proposed method is effective also for containers.

A huge number of inter-connected computers run in data centers. These computers consume large amount of power. Server consolidation with virtualization is a popular method to address this problem. The more computers are consolidated, the more energy is saved. However, highly consolidating, wherein many servers are consolidated into one physical computer, may result in large performance decline. Achieving high consolidation without large performance decline is important. Container-based operating system virtualization is an emerging method for constructing low-overhead virtualized environment. In this work, we focus on Docker, a popular container-based virtualizing system, and investigate its performance, especially performance in highly consolidated environment. First, we compare the performance with and without container-based virtualization, then show that container-based virtualization can provide similar performance to that without virtualization in CPU processing and networking but cannot provide comparable performance in I/O processing with the default setup. Second, we explore the relationship between the number of containers and the obtained performance. We then demonstrate that the I/O performance severely decreases as the number of consolidated servers increase. Third, we discuss the reason why the I/O performance drops largely and applications of highly consolidated servers.

High-performance I/O is essential for big-data analyses. Modern storage systems utilize HDDs and SSDs mainly for achieving large capacity and high performance, respectively. Using an SSD as a cache for accesses to HDDs is one of the promising methods for improving large-scale I/O performance in modern computers. In addition, M.2 is increasing its importance in high-performance I/O processing. In this paper, we investigate the I/O performance of storage systems including M.2 SSD and SSD cache. Our experimental results show that big-data processing performance can improve significantly by using an M.2 SSD cache.

Machine learning is one of the most important fields in recent improvement in big data analysis. Many people apply machine learning for a variety of domains for various purposes, such as classification of opinions. However, the constructed models of machine learning are black boxes. They cannot understand the background reason for their decisions. In many cases, understanding the reasons important. In this paper, we focus on interpretation of models and understanding of decision reasons. First, we introduce the results of an opinions classification of the reviews with Support Vector Machine (SVM). Second, we interpret the model by analyzing weights of the model. Third, we introduce a method for helping to understand the reasons for a decision by SVM by providing a simplified information of the highly weighted words.

When the Great East Japan Earthquake occurred in 2011, the network connectivity was significantly degraded in the wide area due to the multiple network failures as well as the traffic congestion. When the network failures occurred in multiple areas, it was difficult to quickly recognize the entire network situation only using the network traffic monitor system. In our prior works, we found that SNS messages contain the useful information to recognize the big picture of the network failures and proposed the network control system using SNS messages to improve the quickness of the network recovery. As the another critical issue in case of a large-scale disaster, users could not obtain the emergency information due to the network disturbance because the current IP network is operated not being aware of the applications. Thus we propose the application specific traffic control system with failure detection function based on SNS message to prioritize the important application traffic in the event of the large-scale disaster. Based on a series of experiments, this paper shows the effectiveness of a system that detects connection failure based on social information and controls the network bandwidth for each application. Especially, we focus on application specific traffic control experiment. An automatic SDN control is performed with the network virtualization node FLARE having SDN extension capability as well as the network slicing capability. We perform the experiments to determine the type of application based on the traffic and perform bandwidth control for each application using real Internet applications.

When the Great East Japan Earthquake occurred in 2011, the network connectivity was significantly degraded in the wide area due to the multiple network failures as well as the traffic congestion. When the network failures occurred in multiple areas, it was difficult to quickly recognize the entire network situation only using the network traffic monitor system. In our prior works, we found that SNS messages contain the useful information to recognize the big picture of the network failures and proposed the network control system using SNS messages to improve the quickness of the network recovery. As the another critical issue in case of a large-scale disaster, people could not obtain the emergency information due to the network disturbance because the current IP network is operated not being aware of the applications. Thus we propose the specific application traffic control system with failure detection function based on SNS message to prioritize the important application traffic in the event of the large-scale disasters. Based on a series of experiments, this paper shows the effectiveness of a system that detects connection failure based on social information and controls the network bandwidth for each application. Especially, we focus on application specific traffic control experiment. An automatic SDN control is performed with the network virtualization node FLARE having SDN extension capability as well as the network slicing capability. We perform the experiments to determine the type of application based on the traffic and perform bandwidth control for each application using some real Internet applications.

In a virtualized environment, such as a cloud computing environment, guest and host operating systems (OS) run simultaneously. Both of the operating systems have page caches for disk accesses. In such an environment, the second level cache does not work effectively because of negative temporal locality of access and duplicated storing in both the caches. In this paper, we propose a method for increasing the hit ratio of the host operating system page cache. The method monitors pages dropped from the guest operating system page cache and stores the pages into the host operating system cache. After the proposal, we introduce the design and implementation of our system. Our implementation aims to be applicable without modification to the hypervisor, and thus it can be applied for an environment with a proprietary hypervisor. Lastly, we present the evaluation of the hit ratio of the host operating system cache, and then demonstrate that our method can improve the cache hit ratio.