Despite the widespread adoption of electronic medical records (EMR), the variations in format and terminology across institutions hinder inter-institutional comparisons and feature extraction. This paper proposes a demonstration of extracting frequent disease-specific instruction sequences and efficiently visualizing them with statistical insights, e.g. statistical trends, and abnormal inspection result rates from real multi-institutional EMR data. The utility of the developed visualization tool was presented for decision support and improving clinical processes.

As demand grows for event recommendation systems capable of accommodating users with diverse backgrounds and preferences, recent advancements in Large Language Models (LLMs) have attracted significant attention. LLMs offer the ability to flexibly integrate multiple conditions, enabling highly personalized recommendations. However, this flexibility often results in longer prompts, which can degrade response accuracy. Additionally, optimal prompt structures may vary across LLM versions, leading to inconsistent performance.To address these issues, we propose a dynamic prompt optimization method that segments prompts based on LLM evaluations. We conducted experiments using ChatGPT-4o and evaluated the effectiveness of prompt segmentation to verify the feasibility of the proposed method. The experimental results demonstrate that prompt segmentation improves recommendation accuracy but that its effectiveness saturates beyond a certain threshold. These findings suggest that determining the optimal number of prompt segments maximizes output quality while preventing unnecessary segmentation.

Various sensor data from Internet of Things (IoT) devices are expected to be routinely collected, analyzed, and utilized in the cloud. However, the required communication throughput and latency for various services must be maintained when collecting IoT data in mobile environments. IoT communication involves a large amount of small-scale streaming data, necessitating efficient transmission methods tailored to the communication environment. In this study, we investigate the effectiveness of compression processing by varying the publish/subscribe data characteristics and data compression ratios to improve performance. In experiments, the performance of MQTT communication over SINETStream is investigated, varying parameters such as data size, compression algorithm, data characteristics, and data compression ratio. The results show that performance is improved for highly compressible data, and the performance difference becomes more pronounced as data size increases. There is no correlation between compression time and data compression ratio, and the impact of compression time on overall execution time is slight, thereby confirming that selecting appropriate algorithms for each data characteristic and applying compression processing according to the data size effectively improves performance.

Quantum computers can achieve extremely fast computations for certain problems using quantum properties. Due to these factors, research and development in the field of quantum computers have been vigorously pursued. It is crucial to develop a quantum computer that operates accurately and, concurrently, create applications compatible with quantum computing. Quantum computer simulators are tools that represent the behavior of a quantum computer on classical computers, and they are highly conducive to the development of quantum applications that can run on actual quantum computers. On the other hand, quantum computer simulators run on classical computers, which are extremely slow compared to quantum computers. Therefore, it is desirable for the development of quantum applications to be able to fully utilize the computational resources on a classical computer and run at high speed. In this paper, we analyze the performance of the Qiskit Aer simulator, the Qulacs simulator, and mpiQulacs on various types of single servers. We then compare the performance of these quantum computer simulators across different computer environments.

The data collected by Internet of Things (IoT) devices equipped with sensors enable smart home services such as monitoring the elderly, pets, and the indoor environment. Building an IoT system to collect data from individual households in the cloud requires measures to reduce communication latency and the amount of data transferred, and protect privacy. When collecting diverse data in an indoor environment, installing sensors in multiple indoor locations is necessary. However, installing numerous sensors increases costs and makes it difficult to relocate the sensors to obtain the necessary information. In this study, we construct an IoT system for a smart home that collects indoor environmental information using a wheeled mobile robot implemented in a Robot Operating System (ROS) and performs analysis processing in a cloud via an edge server. We attempt to demonstrate the effectiveness of sensor data collection using a robot by developing a prototype system for indoor carbon dioxide concentration monitoring application. We also the performance characteristics of ROS communication between the sensor robot and the edge server, and IoT communication between the edge server and the cloud server under different communication environments to identify technical issues in the smart home.

The use of data collected by edge devices in machine learning, including personal information, has become an important trend in recent years. Most distributed machine learning methods such as Federated Learning aggregate and manage all data or training results on a high-performance edge servers. However, passing users’ personal information to an external server may involve some privacy concerns owing to the risk of information leakage. To address this problem, we consider a distributed machine learning model with excellent privacy protection in which the user can choose not to pass any personal data to the server. In the proposed model, the edge device takes over the training at the edge server and sends only the results for which the user has given permission to the edge server for integration. To validate the effectiveness of the proposed model, we performed experiments on facial image recognition using a Jetson Nano as an edge device. The experimental results confirm that edge devices were able to use personal information in a short period of time, while the edge server was able to obtain more accurate results by integrating several training results. Thus, the results show that the proposed model enables the safe and efficient utilization of data collected by edge devices.

The long-term use of machine learning models can result in degraded performance due to data drift and other factors. We have previously proposed a data selection mechanism for time-series data of machine learning models. When data drift occurs, the models have to learn again using large-scale stream data. Thus, it is important for machine learning algorithms to introduce a mechanism avoiding useless data. This study examines the effect of data selection with an adversarial classifier using synthetic data.

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.

Recently, numerous studies have been conducted on the practical implementation of fully homomorphic encryption (FHE) for large data analysis. Many systems using FHE are cloud-based and use many DRAMs because the progression of encryption or bootstrapping increases the time and space complexities. However, DRAMs in the cloud cannot be effectively used because they are shared by several systems and cost us inefficiently. Here, we combine Solid State Drive (SSDs) with DRAMs for an application using FHE. They have several different architectural advantages from DRAMs: they cost less per one bit and have high parallelism. Our primary objective is to utilize SSDs instead of DRAMs and reduce the cost on the cloud when using an application of FHE. This study uses the HElib library and an application for a privacy-preserving search for genome information. First, we identified the bottleneck of this application for using SSDs effectively. We analyzed the features of the application using Hyper-Threading technology and found that it consumes considerable CPU power. It is highly parallel, and the processing of load/store instructions from DRAMs to CPUs in sequential is the bottleneck. Second, we revealed the bottleneck in detail by the function in HElib library. The ”rotate” function is found to be the bottleneck in it. It must be implemented many times because of the FHE algorithm which cannot compare a ciphertext to another one. Finally, the compatibility of this application using the HElib library with SSDs is determined.

In recent years, with the spread of social networking services (hereinafter referred to as SNS), information on events of various sizes, including local events and ongoing events, has been posted on SNS. Research has already begun to create systems that recommend events for users to attend at specific times and places based on a collection of event information and using user location data. However, for the user, there is a concern regarding privacy that the user's location information or the “like” reaction within SNS to certain events is transmitted to the server as is and used for information analysis. Therefore, in this work, we considered the user's privacy by converting the user's position information into a dummy position using regional mesh information and utilizing it for the inquiry to the server. We examined a method of acquiring event information from Twitter, which is a type of SNS, and searching for event information from a large amount of SNS data after satisfying the geographical constraints on the user side.

What do we need to do to protect our personal information? IoT devices such as smartphones, smart watches, and home appliances are widespread. Encryption is required not only to prevent eavesdropping on communications but also to prevent information leakage from cloud services due to unauthorized access. Therefore, attention is being paid to fully homomorphic encryption (FHE) that allows addition and multiplication between ciphertexts. However, FHE with this convenient function has a drawback that the encryption requires huge volume of calculation and the ciphertext is large. Therefore, if FHE is used on a device with limited computational resources such as an IoT device, the load on the IoT device will be too heavy. In this research, we propose a system that can safely and effectively utilize data without imposing a load on IoT devices. In this system, somewhat homomorphic encryption (SHE), which is a lightweight cryptosystem compared with FHE, is combined with FHE. The results of the experiment confirmed that the load on the IoT device can be reduced to approximately 1/1400 compared to load of the system from previous research.

It is expected that heavy congestion occurs at a time of a severe disaster. In such a situation, traffic priority control in network elements for preferentially transferring information for rescue is important. For achieving this, service identification from IP flows in network elements is required. A method based on deep packet inspection (DPI) of multiple TLS sessions has been proposed for identifying services from IP flows. However, its analysis is extremely CPU time-consuming. In this paper, we propose a new service identification method based on the analysis of Server Name Indication (SNI). SNI is a not-encrypted field in the ClientHello message, which is transferred in the TLS session establishment. We then evaluate the proposed method and show that the proposed methods significantly reduce time to identify without a large decline in the accuracy of identification.

Recent programmable switches allow developers to profoundly optimize network elements. In this study, we introduce a concept of an application switch that supports a network application based on programmable switches. A developer can optimize not only server computers but also application switches to improve the application performance by implementing functions of the network application. We then introduce a method to apply an application switch to a database management system using the transmission control protocol (TCP). The method migrates a TCP connection from a server computer to the application switch when the switch replies to a query. The method reversely migrates it to the server computer again when a query is processed by the server. The application switch manages the TCP sequence and the Ack numbers for these migrations. In the existing work, we implemented a prototype key-value store (KVS) system based on the proposed system without TCP Ack management. We then showed the potential of the method. In this paper, we show the problem of the prototype system of Ack management and propose a method for solving the problem. In addition, we evaluate the resolved method and demonstrate its effectiveness.

With the development of cameras and sensors, and the spread of cloud computing, life logs can be acquired and stored in general households for various services using the logs. However, it is difficult to analyze moving images acquired by a home sensor in real time using machine learning because the data size and the computational complexity are large. New computing paradigm called edge computing or fog computing, which enables distributed computing over edge and cloud, has the possibility to address this issue. The feature vectors are extracted from moving images by preprocessing on the sensor side and the only small feature vectors are sent to the cloud and used for learning. But, it is not clear how accurately we can recognize actions using only the feature vectors in the learning and inferring. We investigate the accuracies of action recognition with various machine learning methods using feature vector information obtained from moving images. We use the pose estimation library OpenPose for detection of the feature vectors and recognize actions using logistic regression, random forest, support vector machine, and neural network (NN) models, general NN and LSTM, as machine learning methods. The experimental results show that it is possible to recognize an action with 80% accuracy or higher when using random forest and neural network models. We also discuss a method to further improve the accuracy based on the experimental results.

When a large-scale disaster occurs and the usual communication infrastructure is disconnected, Internet access becomes impossible, and information cannot be exchanged. However, using edge servers, each terminal can upload and view information. Therefore, we consider an environment construction that enables information exchange by edge servers keeping information synchronized. In this paper, we construct an application that accepts and aggregates requests for items that are required for materials other than those provided at the time of disaster and shares information with goods wholesalers.

Genome sequence search is useful, for example, in clinical applications where a care provider needs to select a treatment option for a patient based on the exact kind of cancer the patient might have. Homomorphic encryption is a desirable technology to be used for this application because it is non-interactive. However, privacy-preserving genome sequence search using homomorphic encryption has been a practical challenge because of scalability issues driven by the depth of computations that need to be supported for privacy-preserving genome sequence search. In this paper, we build off of earlier privacy-preserving genome sequence search results to design, implement and compare two approaches to a client-server style system for privacy-preserving genome sequence search. There is a myriad of options and design trade-offs associated with the application of homomorphic encryption in this domain driven, for example, by choices in data encoding, scheme selection, and even encryption software library. We particularly focus on the use of the BGV and BFV homomorphic encryption schemes provided by the HElib and PALISADE open-source homomorphic encryption software libraries. Our results show that using the BFV-based approach in PALISADE provides optimal results for this application over our sample data.

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.

SNS websites and messages have become a regular part of daily life. However, some large-scale disasters such as earthquakes could damage network cables and equipment and cause network failures. Despite the assistance of information related to disasters and support that is effective for controlling the situation and offering help, it is possible that frequently used SNS messages and websites will become unreachable in a disaster area. Moreover, in a disaster area, the network could break into several sub-networks because of the damage of the backbone network. It is possible that the sub-networks can provide network services in a smaller area range without connecting to the Internet. Therefore, this paper proposes a decentralized SNS system that could solve two major problems during large-scale disasters. First, it describes a stable and continuous SNS network for the affected population and helpers without the influence of networks in the disaster region. Second, it proposes a reliable method to share synchronized data across the affected areas. In this case, a reliable method to control connections among shelters, whether newly built shelters or shelters that lost connection during the disaster, should also be considered. This paper shows the effectiveness of the proposed decentralized SNS system using XMPP to provide SNS for refugees and helpers in each shelter under an emergency of network failure. The connection control among shelters has also been examined and was designed to provide secure connections and data synchronization for this decentralized SNS system.

The spread of various sensors and the development of cloud computing technologies enable the accumulation and use of many live logs in ordinary homes. In addition, deep learning technologies have been widely used for image and speech recognition processing. However, a key issue for deep learning is heavy processing loads. To operate a service that utilizes sensor data, those data are transmitted from sensors in ordinary homes to a cloud and analyzed in the cloud. However, services that involve moving image analysis require large amounts of data to be transferred continuously and high computing power for the analysis; hence, it is difficult to process them in real time in the cloud using a conventional stream data processing framework. First, we perform preliminary experiments using Apache Spark [3] (hereinafter called Spark), which is a representative cluster computing platform that is designed to be fast and versatile, and Ray [4] , which is a distributed execution framework. We investigate the characteristics of their distributed recognition processing and demonstrate that Ray enables scalable distributed processing. Next, We implement a prototype system of the proposed distributed stream processing infrastructure using Ray and Apache Kafka [1] (hereinafter called Kafka), which is a distributed messaging system, and demonstrate its performance.

Many services, such as email, video sharing, and social networking service (SNS), are provided on the Internet. Service identification from given flows is important for various purposes. For example, a severe congestion occurs in disasters and priority control is required for transmitting important information, such as requests for rescues, in that case. Identification of the service of a traffic in network elements achieves this control. The most simple way to identify is that based on IP addresses and port numbers. However, the accuracy of this way is not sufficient. A method for identifying service based on analyzing multiple TSL sessions without using IP addresses and port numbers was proposed. This method clusters TLS sessions according to the 2-gram frequencies of unencrypted parts, which are the fields in handshake messages transmitted at session establishing. However, the existing work did not discuss the effect of each field of the TLS protocol. In this paper, we analyze the ability to cluster of each field. We investigate the ability to cluster sessions of all the unencrypted fields of the handshake messages. We then reveal that some fields do not have the ability. We discuss methods for improving the existing method based on these finding.

As the use of various sensors and cloud computing technologies has spread, many life-log analysis applications for safety services for the elderly and children have been developed. However, it is difficult to perform real-time large data processing in clouds due to the computational complexity of the analysis because efficient deployment schemes of streaming computing components over cloud resources have not been well-investigated. In this study, we propose a video analysis framework that collects videos from multiple cameras and analyzes them using Apache Kafka and Apache Spark Streaming. We first investigate the data transfer performance of Apache Kafka and examine efficient cluster configuration and parameter settings. We then apply this configuration to the proposed framework and measure the data analysis throughput. The experimental results show that the overall throughput varies depending on the number of broker nodes that store data, the number of topic partitions of data, and the number of nodes that conduct analysis processing. In addition, it is confirmed that the number of cores is needed to consider for the efficient cluster configuration, and that the network bandwidth between the nodes becomes a bottleneck as the amount of data and the number of components increase.

The Spread of cameras and sensors and cloud technologies enable us to obtain life logs at ordinary homes and transmit the captured data to a cloud for life log analysis. However, the amount of processing for video data analysis in a cloud drastically increases when a very large number of homes send data to the cloud. In this research, we aim to improve the efficiency of distributed video data analysis processing by using the parallel deep learning framework Chainer [2] and the distribution processing platform Apache Spark [1] (Spark). In this paper, we construct a Spark cluster and investigate the performance of parallel data processing using Spark varying parameter settings.

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.

Current network switches cannot be programmed and flexibly controlled. Then, developers of a data center application system, which is composed of software and computers connected with a network, are not able to optimize behavior of network switches on which the application is running. On the other hand, Deeply Programmable Network (DPN) switches can completely analyze packet payloads and be profoundly programmed. In our previous work, we introduced an application switch based on DPN. The switch was able to be deeply programed and developers could implement a part of functions of a data center application in the switch. The switch deeply analyzed packets, which is called Deep Packet Inspection (DPI), and provided some functions of the application in the switch. However, the switch did not manage connection and not support communication with TCP. In this paper, we proposed a method for constructing an application switch supporting TCP based communication. The method analyzes IP headers, TCP headers, and payloads of packets. When the switch detects a request which the switch supports, the switch replies according to its TCP session. We then introduce our implementation and evaluate performance of our application switch. Our evaluation has demonstrated that our switch has been able to improve performance of the data center applications.

Bigdata processing platform, such as cloud computing using container-based virtualization, enabled large-scale sensor data analyzes. In addition, smartphones have been popular, estimation of a lot of information, such as user's behavior, by analyzing its sensor data has become possible. In this study, we focus on the estimation of whether the user is tall or short as the first study of estimating the user's height from the data of the triaxial acceleration sensor during walking. We propose methods for estimating the height of the user from the sensor data using machine learning. We then show its effectiveness by performance evaluation.

An outsourcing system for calculating the statistics of genome data has been proposed as a method for utilizing genomic data in bioinformatics research. Individual genome data need to be encrypted from the viewpoint of privacy protection. In previous research, a client/server system was proposed for the string search of genome sequences using fully homomorphic encryption. The authors also improved the query and calculation algorithms for taking more advanced statistics of the genomic data for use in the future. However, calculations using fully homomorphic encryption are highly complex. In this study, we proposed the implementation of a master/worker decentralized system on the server side of a string search system for the genome using fully homomorphic encryption. We propose a decentralized system that is operated in a cloud computing environment.

Large scale applications in data centers are composed of computers connected with a network. Traditional network switches do not perform routing based on packet contents. Thus, packets cannot be transmitted to the optimal computer for the request which is written in the packet payload. On the other hand, Deeply Programmable Network (DPN) switches can completely analyze packet payloads and perform routing based on their contents. In this paper, we focus on packet routing based on Deep Packet Inspection (DPI), i.e. analyzing packet payload, using DPN switches, and discuss performance of applications on a network using application aware DPN switches. We introduce application aware network and propose a method for improving application performance with DPN switch support. We evaluate our method with the several applications and show that our method can increase application performance.

The spread of various sensors and Cloud technologies has made it easy to acquire life-logs and accumulate data. As a result, many life-log analysis applications, which transfer data from sensors, especially cameras to a Cloud and analyze them in the Cloud, have been developed. Cameras with a server function called network cameras have become cheap and readily available for security services and the monitoring of pets and children from remote locations. In these services, raw data from sensors, including cameras, are generally transferred to a Cloud and processed there. However, it is difficult to transfer raw data from sensors to a Cloud because of the limitation of network bandwidth between sensors and a Cloud and privacy issues caused by sending raw sensor data to a Cloud.

This paper shows the effectiveness of the system that controls the network route based on social information. An automatic route control is performed with the network virtualization node FLARE that uses VLAN coupling between each base of the wide area network test bed called JGN-X as a platform.

To achieve network availability in disaster situations, we propose the traffic control system based on SNS information. In order to utilize the data of packet payload obtained from SNS, Deeply Programmable Network (DPN) is needed. In this paper, we implement and evaluate the proposed method with FLARE switch and achieve more flexible control of the network.

An incentive problem of ad hoc networks still remains for relaying frames from other nodes. By utilizing the Social Network System such as Twitter and Facebook, it is possible to estimate the degree of intimacy of friendship links not only to direct friends but also indirect friends such as friends-of-friends. If we provide links of ad-hoc network in relation to direct/indirect friends instead of other incentive mechanisms, ad-hoc network can obtain infinite potentials to connect all nodes efficiently and cover wide areas by short range communication such as IEEE802 wireless LANs. This paper is devoted to discuss social network based (real) ad-hoc network concept and evaluation of its connection probability performance.

In this paper, QoS characteristics such as TCP throughput is investigated for densely deployed mobile wireless LANs (WLANs). Factors affecting throughput characteristics are discussed and evaluated by using real machines such as smartphones and portable APs. In IEEE 802.11 WLANs, a rate adaptation mechanism controls the transmission rate and one of the dominant factors for QoS. In order to understand the behavior of the rate adaptation control, 1 to 18 sets mobile WLANs are examined under different parameters. Since a behavior of the rate adaptation control is a vender specific one and it strongly depends on interference, signal strength and etc., the real terminals such as smartphones were used in the experiments. Performance anomaly drastically reduces the throughput not only in the WLAN which has a terminal with low transmission rate but also in the neighboring WLANs that share the same channel. In order to avoid unnecessary transmission rate degradation by the rate adaptation control, Context Aware multi Rate Control (CARC) is proposed and evaluated. In CARC, Turning the rate adaptation control on/off is controlled according to a context, for example, signal strength. The evaluation results show that the CARC can be cost-effectively implemented and improves the throughput performance of whole WLANs by 3.5 times than that without the application of CARC.

A social media message that is forwarded (e.g., Retweeted in Twitter) by many users may diffuse widely. Such a 'viral' message may have a major impact in the real world. If the content maligns a company, the company must respond as quickly as possible to defend its brand image. Timely knowledge about what kinds of information are diffusing in social media is thus quite important. We have developed a system that enables real-time analysis of information diffusion in social media. We defined typical data access patterns and created a special view for complicated queries such as top-k ranking queries to improve query performance. The time window is customized by estimating diffusion extinction, which enables retention of diffusion data in the system as long as it is being retweeted frequently. Evaluation showed that the processing of simple access patterns was extremely fast, that the special view effectively improved performance, and that the time-window customization improved accuracy.

This paper investigates Quality of Service (QoS) of the personal mobile wireless LANs (m-WLANs). The situations in the m-WLANs differs from the situations in the normal use of WLANs; the access point (AP) and the associated terminals (TEs) are in proximity. In the m-WLANs, the capture effect (CE) significantly affects on the throughput performance. To measure the impact on the QoS by the CE, the experimental study considering the interference from other power sources (APs and TEs) are required. However, since it is difficult to understand the detailed relationships between the QoS factors, the analytical calculations were also performed. With the experimental and analytical results, we demonstrated that the auto rate fallback algorithm of WLAN causes degradation of the QoS performance. We propose two transmission rates controlling schemes to improve the QoS performance.

Distributed file systems, which manage large amounts of data over multiple commercially available machines, have attracted attention as a management and processing system for big data applications. A distributed file system consists of multiple data nodes and provides reliability and availability by holding multiple replicas of data. Due to system failure or maintenance, a data node may be removed from the system and the data blocks the removed data node held are lost. If data blocks are missing, the access load of the other data nodes that hold the lost data blocks increases, and as a result the performance of data processing over the distributed file system decreases. Therefore, replica reconstruction is an important issue to reallocate the missing data blocks in order to prevent such performance degradation. The Hadoop Distributed File System (HDFS) is a widely used distributed file system. In the HDFS replica reconstruction process, source and destination data nodes for replication are selected randomly. We found that this replica reconstruction scheme is inefficient because data transfer is biased. Therefore, we propose two more effective replica reconstruction schemes that aim to balance the workloads of replication processes. Our proposed replication scheduling strategy assumes that nodes are arranged in a ring and data blocks are transferred based on this one-directional ring structure to minimize the difference of the amount of transfer data of each node. Based on this strategy, we propose two replica reconstruction schemes, an optimization scheme and a heuristic scheme. We have implemented the proposed schemes in HDFS and evaluated them on an actual HDFS cluster. From the experiments, we confirm that the replica reconstruction throughput of the proposed schemes show a 45% improvement compared to that of the default scheme. We also verify that the heuristic scheme is effective because it shows performance comparable to the optimization scheme and can be more scalable than the optimization scheme.

Due to the rapid development of sensor devices, many types of multimedia sensor data can be collected and analyzed to develop useful multimedia applications such as human activity recognition. Our objective of this research work is to study and analyze the impact of the wireless multimedia network quality on human activity recognition accuracy in an end-to-end networked multimedia system environment - Ocha House. We utilized the new improved Generalized Time Warping (GTW) algorithm and its variants for temporally aligning multi-modal sequences from multiple subjects performing similar human activities. In Ocha House, we captured the human activity data with camera sensors which are transmitted via wireless network for server processing and analysis. We evaluated the human activity recognition accuracy using GTW and its variants by taking into consideration of the wireless network data packet loss phenomenon due to common wireless network protocol behaviors as a result of network flow control and congestion management. We discovered that GTW and its variants are sensitive to wireless network packet loss and observed interesting characteristics resulting in deteriorated human activity recognition accuracy - when wireless network packet loss rate is nominal, the human activity recognition accuracy decreases dramatically.

This paper presents a study of mobile wireless LAN (WLAN) hotspots which are used to provide cellular-WiFi tethering service to personal devices. A dense deployment scenario for fixed and mobile WLAN is described and potential performance problems due to interference are identified. An analytical model for coexisting fixed and mobile WLAN hotspots with heterogeneous traffic is presented. The model is used to evaluate the performance of a mobile WLAN as it transits through a set of densely deployed fixed access points (APs), and performance problems due to lack of frequency coordination are identified. An adaptive channel assignment (ACA) scheme for improving mobile AP performance is proposed and evaluated. It is shown that significant performance gains can be achieved with ACA with maximum absolute and percentage throughput gains up to 1.24 Mbps and 42.8% respectively. We also show that setting the scanning interval in ACA requires consideration of the speed at which the mobile WLAN is moving in order to compensate for the throughput losses during channel scanning.

In recent years, due to digital convergence mobile devices that connect to the Internet, e.g. smartphones and tablets, have grown exponentially. Smartphones, which replaces the PDA or the cellular phone, has come to use much bandwidth like a general computer. However, it is a problem that the throughput of wireless communication is deteriorated by traffic congestion when multiple smartphones share one single access point (AP). TCP congestion control plays a critical role in the end-to-end connection in order not to overflow the traffic although most of the bottleneck of network bandwidth is the wireless part. In this paper, we present a middleware for cooperative transmission control on smartphones. The middleware exchanges information among the other smartphones in the same wireless LAN(WLAN) and helps smartphone adjust the congestion control to surrounding environment. As the result, both communication throughput and the fairness for bandwidth utilization are improved with the cooperative control. We implemented the middleware on the Android platform and evaluated it with real devices to show the effectiveness.

Mobile ad hoc network (MANET) is an autonomous distributed network which is constructed only with gathered nodes on the spot. In MANET, wide area communication on wireless networks can be realized because a node relays other node's packets by adopting a multi-hop routing protocol. Security is an indispensable issue in MANET because an unknown node may join the network easily. However, the demand for the level of authentication and its response time is different depending on the environment and applications. Thus, we focus on real-time control of the security mechanism for MANET. In this paper, we have established multi-hop communication environment and introduced methods to control secure connections by applying IPsec as an encryption scheme, on top of the general-purpose operating systems with enabled preemption for interruption. Under such conditions, we have measured response time of CPU with or without load of other applications. We have evaluated the influence of streaming as an application load and investigated on a security realization method for MANET.

Since IP-SAN allows us to reduce initial costs and management costs of storage by using IP networks, iSCSI, common protocol in IP-SAN, is becoming more important. Although one of the key issues for iSCSI is a security measure to access remote storage over IP networks, this is not necessarily established yet. iSCSI can employ IPsec that offers a function of strong encryption. However, the IPsec encryption processing degrades throughput of storage access and increases its CPU load. In addition, it is difficult to execute encryption processing efficiently, because IPsec layer is located in a lower-level. In this paper, we implemented a middleware system of executing encryption processing in the upper-layer for the communication using iSCSI securely and effectively. Moreover, we simulated the idea of optimization of encryption processing and evaluated its performance by running parallel processes in a long-latency network. As a result, optimization method of encryption processing is more efficient than that of IPsec in a long-latency network

Personal computer/workstation (PC/WS) clusters have come to be studied intensively in the field of parallel and distributed computing. From the viewpoint of applications, data intensive applications including data mining and ad-hoc query processing in databases are considered very important for massively parallel processors, in addition to the conventional scientific calculation. Thus, investigating the feasibility of such applications on a PC cluster is meaningful. A PC cluster connected with a storage area network (SAN) is built and evaluated with a data mining application. In the case of a SAN-connected cluster, each node can access all shared disks directly without using a LAN; thus, SAN-connected clusters achieve much better performance than LAN-connected clusters for disk-to-disk copy operations. However, if a lot of nodes access the same shared disk simultaneously, application performance degrades due to the I/O-bottleneck. A runtime data declustering method, in which data is declustered to several other disks dynamically during the execution of the application, is proposed to resolve this problem.

Personal computer/workstation (PC/WS) clusters have become a hot research topic in the field of parallel and distributed computing. They are considered to play an important role as a large scale computer system, such as large server sites and/or high performance parallel computers, because of their good scalability and cost performance ratio. In the viewpoint of applications, data intensive applications such as data mining and ad-hoc query processing in databases are considered very important for massively parallel processors, in addition to the conventional scientific calculation. Thus, investigating the feasibility of such applications on a PC cluster is meaningful. A PC cluster connected with a storage area network (SAN) is built and evaluated. For disk-to-disk copy operation, SAN clusters are much better than LAN clusters. A data mining application is implemented on the cluster. This application requires iterative scans of shared disks, which degrade the execution performance due to I/O-bottleneck. In order to resolve the problem, a dynamic data copy method is proposed and evaluated. This method prevents the performance degradation caused by shared disk bottleneck in SAN clusters.

Personal computer/Workstation (PC/WS) clusters are promising candidates for future high performance computers, because of their good scalability and cost performance ratio. Data intensive applications, such as data mining and ad hoc query processing in databases, are considered very important for massively parallel processors, as well as conventional scientific calculations. Thus, investigating the feasibility of data intensive applications on a PC cluster is meaningful. Association rule mining, one of the best-known problems in data mining, differs from conventional scientific calculations in its usage of main memory. It allocates many small data areas in main memory, and the number of those areas suddenly grows enormously during execution. As a result, the contents of memory must be swapped out if the requirement for memory space exceeds the real memory size. However, because the size of each data area is rather small and the elements are accessed almost at random, swapping out to a storage device must degrade the performance severely. In this paper, we investigate the feasibility of using available remote nodes' memory as a swap area when application execution nodes need to swap out their real memory contents during the execution of parallel data mining on PC clusters. We report our experiments in which application execution nodes acquire extra memory dynamically from several available remote nodes through an ATM network. A method of remote memory utilization with remote update operations is proposed and evaluated. The experimental results on our PC cluster show that the proposed method is expected to be considerably better than using hard disks as a swapping device. The dynamic decision mechanism for remote memory availability and the migration operations are also evaluated.

PC clusters have been studied intensively for next-generation large scale parallel computers. ATM technology is a strong candidate as a de facto standard of high speed communication networks. Therefore an ATM connected PC cluster is a very promising platform from the cost/performance point of view, as a future high performance computing environment. An ATM connected PC cluster consisting of 100 PCs is reported, and characteristics of a transport layer protocol for the PC cluster are evaluated. Point-to-point communication performance is measured and discussed when a TCP window size parameter is changed. Retransmission caused by cell loss at the ATM switch is analyzed, and parameters of the retransmission mechanism suitable for parallel processing on the large scale PC cluster are clarified. From the viewpoint of applications, data intensive applications such as data mining and ad-hoc query processing in databases are considered to be very important for massively parallel processors, in addition to conventional scientific calculations. Thus, investigating the feasibility of such applications on an ATM connected PC cluster is quite meaningful. Parallel data mining is implemented and evaluated on the cluster. The default TCP protocol cannot provide good performance, since a lot of collisions happen during all-to-all multicasting executed on the large scale PC cluster. Using TCP parameters according to the proposed optimization, sufficient performance improvement is achieved for parallel data mining on 100 PCs.

We developed a PC cluster system consists of 100 PCs. Each PC employs the 200MHz Pentium Pro CPU and is connected with others through an ATM switch. We picked up two kinds of data intensive applications. One is decision support query processing. And the other is data mining, specifically, association rule mining.As a high speed network, ATM technology has recently come to be a de facto standard. While other high performance network standards are also available, ATM networks are widely used from local area to widely distributed environments. One of the problems of the ATM networks is its high latencies, in contrast to their higher bandwidths. This is usually considered a serious flaw of ATM in composing high performance massively parallel processors. However, applications such as large scale database analyses are insensitive to the communication latency, requiring only the bandwidth.On the other hand, the performance of personal computers is increasing rapidly these days while the prices of PCs continue to fall at a much faster rate than workstations'. The 200MHz Pentium Pro CPU is competitive in integer performance to the processor chips found in workstations. Although it is still weak at floating point operations, they are not frequently used in database applications.Thus, by combining PCs and ATM switches we can construct a large scale parallel platform very easily and very inexpensively. In this paper, we examine how such a system can help the data warehouse processing, which currently runs on expensive high-end mainframes and/or workstation servers.In our first experiment, we used the most complex query of the standard benchmark, TPC-D, on a 100 GB database to evaluate the system compared with commercial parallel systems. Our PC cluster exhibited much higher performance compared with those in current TPC benchmark reports. Second, we parallelized association rule mining and ran large scale data mining on the PC cluster. Sufficiently high linearity was obtained. Thus we believe that such commodity based PC clusters will play a very important role in large scale database processing.

The information retrieval systems on a wide area distributed network, such as the World-Wide Web (WWW), become popular among the extremely large number of users. The caching proxy has an important role in these systems for improving their accessibility and serviceability. The caching proxy mechanism is discussed in this paper. First, the role and structure of the caching proxy is explained, and two major problems of existing systems are pointed out. Our solution to overcome these problems is proposed next. Changing a file size by controlling a quality level of cached multimedia data is proposed as a measure to overcome one problem. As a solution to the other problem, making a cluster among neighboring caching proxies, using hyperlink information, is proposed. Finally, an improved caching proxy mechanism based upon these ideas is shown.

To realize functionally distributed computing in a wide area distributed environment, distributed shared memory (DSM) is an attractive option due to the simplicity and flexibility in software programming. DSM has mainly been studied in a local environment. In a widely distributed environment, the latency of communication impacts system performance, even if a high bandwidth network is used. DSM models in a widely distributed environment are discussed and evaluated in this paper. First, two existing DSM models are examined: shared virtual memory and replicated shared memory. Next, an innovative replicated shared memory model, which uses internal machine memory, is proposed. A prototype of this model using multi-thread programming was implemented on multi-CPU SPARCstations. These DSM models are compared with SCRAMNet, whose mechanism is based on replicated shared memory. Results from this evaluation show the superiority of the replicated shared memory compared to shared virtual memory when the length of the network is large. While replicated shared memory using external memory is influenced by the ratio of local and global accesses, replicated shared memory using internal machine memory is suitable for a wide variety of cases. The replicated shared memory model is considered to be suitable particularly for applications which impose real time operation in a widely distributed environment, since some latency hiding techniques such as context switching or data prefetching are not effective for real time demands.