Such home networking technologies as UPnP have become common for users to connect their devices at home. In this paper, we propose dynamic overlay topology optimizing content search (DOTOCS) that enables flexible content searches while optimizing P2P overlay networks. DOTOCS allows nodes to change their logical links to reach other nodes that have frequent data transfers in fewer hops, providing users with a better chance of finding their desired contents in quicker downloading time. For this reason, DOTOCS is useful to establish efficient connections among UPnP gateways on unstructured P2P networks for content sharing. We implemented DOTOCS using a P2P simulator to evaluate and measure its performance. The simulation results proved that DOTOCS increases search performance by 30 to 40% compared with a simple content search method that omits optimization.
Commercial cloud services hold great promise as excellent and economical computing resources for running ever-expanding scientific computing workflows. However, there is no easy way to determine which computing resource will be able to run a scientific computing workflow at the lowest cost, so selecting from among the many computing resources available is difficult for scientists. This research aims to provide a clear standard for selecting computing resources for running scientific computing workflows. This research focuses on the issue of usage cost. The issue is how to select instances that will satisfy computing performance requirements and have the lowest user cost when executing a scientific workflow on a commercial cloud service.
Recently, many kinds of embedded devices have been connected through networks aimed toward a ubiquitous network, and such devices are being used to construct distributed multimedia systems. Peer-to-peer (P2P) communication architecture, which is different from legacy client-server communication architecture, embodies one of the key technical concepts of the next-generation network technology. We have proposed a new network framework named SONICA, Service Oriented Network Interoperability for Components Adaptation. This framework is based on the distributed multiple server approach, in which each embedded device has a server function and a client acquires a required function by consolidating some of the distributed server functions. In this paper, we analyze the requirements of ubiquitous networks and apply P2P architecture based on SIP to SONICA. We also evaluate the performance of device control with UPnP and Jini by using implemented systems.
There are high expectations for commercial cloud services as an economical computation resource when executing scientific computing workflows, for which the computation is increasing on a daily basis.However, no method has been developed for determining whether a scientific computing workflow can be executed at a low usage cost, and thus scientists have difficulty in selecting from the diverse range of computational resources.The aim of this study is to provide clear criteria for selecting a computational resource while executing a scientific computing workflow.This study focuses on the performance of application execution for one such commercial cloud service, Amazon EC2, and proposes a method for selecting the optimal resource showing high similarity to a target application in execution time and usage cost.The novelty of this study is its approach of employing application similarity in resource selection, which enables us to apply our method to unknown applications.The contributions of this work include (1) formularizing performance values of computational resources, as well as similarity values of applications, and (2) demonstrating the effectiveness of using these values for resource selection.
This paper proposes a new internet of things platform to share function of devices to users that meet flexible conditions. Internet of things is an architecture to retrieve information of content and uses functions of devices. In previous studies, internet of things' platforms focus mainly on sensing environments by connecting sensors and can only set the sharing condition from either private or public. However, most devices belong to person, enterprise, industry and infrastructure require a platform to set sharing conditions by owner relationship and device characteristics. In this study, this requirement for the internet of things platform is considered. To meet the requirement, a new platform using social network information is implemented.
Recent advances in multimedia network systems have led to the development of a new generation of applications that associate the use of various multimedia objects. The connection of audio and video devices through high speed multimedia pervasive networks will make the use of these systems more convenient. For example, many home appliances, such as video cameras, display monitors, video recorders and audio systems, will be equipped with a communication interface in the near future. Recently some platforms (i.e. UPnP, Jini, HAVi and so on) have been proposed for constructing multimedia pervasive networks, however, there are still some issues to be solved before we can achieve various services that connect different equipment via the network. In case of UPnP based on SOAP, it is necessary to standardize a way to control any feature of a new (undefined) device connected to the network and update the software in the controller device. Meanwhile, Jini based on RMI technology, or HAVi developed for intelligent AV equipments, has a mechanism to upload software module from a new device connected to the network to the controller device. Although standardizing the new devices' features is not needed, HAVi devices require high CPU power and large memory. Considering that the targets of these home appliances are embedded systems, this situation raises issues of software and hardware complexity, cost, power consumption, and so on. In this study, we developed SONICA, a service oriented network architecture for the control and management of home appliances to address the issues described above.
The main current method for obtaining traffic volume data in surveys involves survey personnel using counters. However, at least two persons are required for each lane to obtain a high level of accuracy, because the counting is done in real time. As a result, work cost is high and reducing cost while maintaining the same level of accuracy is a problem. As a cost reduction measure that addresses that problem, the use of crowdsourcing for measuring traffic is proposed and the usefulness of that measure is examined. Specifically, an evaluation of accuracy when data is acquired by an unspecified large number of workers shows that the proposed method can reduce survey cost while maintaining the same accuracy as the existing method by eliminating unproductive time. The results demonstrate the feasibility of applying crowdsourcing, which uses human processing capability as opposed to big data processing.
Universal Plug and Play (UPnP) 1 has been proposed by Microsoft for realization of interactions among home appliances (e.g., TVs, air conditioners, and lighting devices) and information devices (e.g., PCs and PDAs) on a home local network. UPnP assumes to be used on a local network. Difficulties exist when establishing a communication link between two local networks across the Internet. In this paper, we propose a new device interaction method across the Internet with web service technology. We implement prototype system and make evaluation about latency.
