Satoru Noguchi

Proposal and Evaluation of a Dynamic Resource Allocation Method Based on the Load of VMs on IaaS

Recently, Cloud computing has emerged as a new computing paradigm on the Internet. Cloud computing facilitates flexible and efficient computer resource management via virtualization technologies at anytime and from anywhere, so that users can add and/or delete IT resources. Users can set up and boot the required resources and they have to pay only for the required resources. However, they have to spend a considerable amount of time and money to design, set up, boot, and monitor their resources. Thus, in the future, providing a mechanism for efficient resource assignment and management will be an important objective of cloud computing. In this paper, we propose a dynamic resource allocation method based on the load of VMs on IaaS, abbreviated as DAIaS. This method enables users to dynamically add and/or delete one or more instances on the basis of the load and the conditions specified by the user. We implement a prototype to evaluate the effectiveness and efficiency of DAIaS. Furthermore, we perform an experiment to extract the prototype on a real cloud service, namely, Amazon EC2.

Location-aware service discovery on IPv6 GeoNetworking for VANET

Service discovery is an essential component for applications in vehicular communication systems. While there have been numerous service discovery protocols dedicated to a local network, mobile ad-hoc networks and the Internet, in vehicular communication systems, applications pose additional requirements; They need to discover services according to geographical position. In this paper, we propose a location-aware service discovery mechanism for Vehicular Ad-hoc NETwork (VANET). The proposed mechanism exploits IPv6 multicast on top of IPv6 GeoNetworking specified by the GeoNet project. Thanks to the GeoBroadcast mechanism, it efficiently propagates service discovery messages to a subset of nodes inside a relevant geographical area with encapsulating IPv6 multicast packets. We implemented the mechanism using CarGeo6, an open source implementation of IPv6 GeoNetworking. Our real field evaluation shows the system can discover services with low latency and low bandwidth usage in VANETs.

Vehicle clustering algorithm for sharing information on traffic congestion

We present a method for clustering vehicles that are in the same congested traffic flow. Our goal is to provide a mechanism for sharing information between vehicles in the same situation to ease traffic congestion in urban areas. Because the most accurate source of information about the congested traffic flow is the vehicle at the head of the traffic flow, it first needs to be identified. To do so, we adapt a clustering algorithm by trajectory abstraction. By grouping the vehicles that have the same or similar trajectory into the same cluster, the vehicle at the tail of the traffic flow can discover the vehicle at the head of the traffic flow. Moreover, we adapt an abstracted trajectory representation to compensate for the error in GPS information. Simulation results show that the proposed algorithm provides a higher rate of correctness than existing commonly used clustering algorithms.

TrafficCam: Sharing traffic information based on dynamic IPv6 multicast group assignment using smartphone sensors

Exchanging traffic information among nearby vehicles is one of the communication scenarios in the field of intelligent transport systems (ITS). An important question in this scenario is “How can traffic information be delivered to only one subset of nearby vehicles in the same road traffic situation?”, for example, to only congested traffic in a certain lane. In this demonstration, we propose a dynamic IPv6 multicast group assignment mechanism using GPS, an accelerometer and a magnetometer embedded in smartphones. Our mechanism enables each vehicle to find the relevant vehicles, and to compose a temporary IPv6 multicast address. Our mechanism can deliver traffic information to a group of vehicles in a particular situation. This demonstration, carried out on virtual traffic congestion, shows an information-sharing scenario using a prototype of the proposed mechanism, implemented as an Android application called TrafficCam. Our application demonstrates the usefulness of the proposed mechanism.

Performance Analysis of Mobile Publish-Subscribe Service Discovery on IPv6 over GeoNetworking

Distributed applications for intelligent transport systems need to discover services according to geographical position not only within vehicular ad-hoc network (VANET) but also through the Internet. While wide variety of service discovery technologies have been proposed, rapidly locating an appropriate service among huge number of mobile nodes is still challenging issue. Furthermore, managing such large number of mobile services in the Internet raises scalability issue due to mobility of vehicles. In this paper, we propose a wide-area, geographical mobile service discovery system by adapting the publish-subscribe scheme to vehicular communication systems based on IPv6 over Geo Networking. The proposed mechanism integrates a hierarchical publish-subscribe architecture into interconnected VANETs on the standardized Geo Networking protocol. Furthermore we present a cross-layer publication management mechanism with exploiting the IPv6 router advertisement packets. Thanks to the geographical routing, our system efficiently manages the large number of mobile services in the edge of the considered network. We study the basic control overhead of our system by presenting a simple analytical model for I2V and V2I2V discovery scenarios.

Efficient Data Management using the Session Log in DHT and its Evaluation

Distributed Hash Table (DHT) is a specialized distributed system that aims to lookup hashed identifiers efficiently in order to route messages to and from the corresponding nodes and objects. However DHTs are mechanisms sensitive to churn (the continuous and interleaved process of arrival / departure of nodes). As a result, they often cause loss of stored objects and failures in the lookup, a serious issue for all overlay networks. In this paper, we propose a dynamic replica location method using the session log in DHT. Our method can determine dynamically the effective number of replicas on session running periods among anticipate nodes. We also implemented a prototype based on an existing DHT routing algorithm, and we evaluated the performance of the prototype. Our proposed method was shown to reduce data loss and the cost for locating replication.

Koshien-Cloud: Operations of Distributed Cloud as a Large Scale Web Contents Distribution Platform

The cloud computing paradigm is now widely used thanks to its scalable on-demand resource management techniques. Since one of the most important challenges of cloud computing services is to stably provide their consumers with computing resources, it is necessary to investigate the performance of cloud computing services in actual operations. To study the realistic performance of cloud computing services, we propose in this paper deploying and operating a distributed cloud computing system as a web-based, nationwide image delivery service for the annual baseball championship game in Japan. Measurement results show that the proposed system has worked stably, though it raised some performance issues related to the highly loaded VM.

A Divide and Merge Method for Sensor Data Processing on Large-Scale Publish/Subscribe Systems

As sensor-networking technologies have rapidly been developed, various sensor data have become available thanks to Publish/Subscribe mechanisms. Aggregating and combining such different types of sensor data, such as the amount of rainfall and water levels of rivers, can help to develop more valuable applications. However, these processes may cause load concentrations on a particular processing node, which consequently may cause a scalability issue. To handle this issue, we consider a division strategy that is appropriate for large sensor networks as well as providing a data processing mechanism. This system enhances the application fields in ubiquitous sensing environments. In this paper, we propose a highly scalable sensor data processing mechanism on the basis of a content-based network. Our mechanism has a load distribution mechanism that dynamically divides and moves subscriptions so that our system can efficiently avoid an excessive load on a particular processing node. The performance evaluation of our proposed system shows that the load distribution mechanism works well and has high scalability.

Design and field evaluation of geographical location-aware service discovery on IPv6 GeoNetworking for VANET

Service discovery is an essential component for distributed mobile applications in vehicular communication systems. While there have been numerous service discovery protocols, applications for vehicular communication systems pose additional requirements: discover services according to geographical position inside dynamic mobile environments. In this article, we propose a geographical location aware service discovery mechanism for vehicular ad-hoc networks (VANETs). The proposed mechanism exploits an IPv6 multicast service discovery protocol on IPv6 GeoNet-working specified by the GeoNet project. Thanks to the GeoBroadcast mechanism, it efficiently propagates service discovery messages to a subset of nodes inside a relevant geographical area with encapsulating IPv6 multicast packets. We implemented the proposed mechanism to the ns-3 network simulator, furthermore we integrated the prototype system using CarGeo6, an open source implementation of IPv6 GeoNetworking, with openSLP. Our simulation and real field evaluation results show the system can discover services with low latency and low bandwidth usage in VANETs even via multi-hop.

A Distributed Publish/Subscribe System for Large Scale Sensor Networks

As sensor and wireless network technologies are developed, it is expected that we will be able to gather lots of sensor data from various sensors on globe. Combining/ aggregating these sensor data, many useful applications could be developed. However, it is difficult to aggregate sensor data and secure scalability, because in some case too much sensor data concentrate on some specific nodes. Then, we add data processing components onto content-based networks, considering division and reassign of processing components. Content-based networks provide publish/ subscribe system in distributed environment and appropriate for notifications and alerts. If content-based networks process published data and create meaningful information, this system enhance the application fields in ubiquitous sensing environment. In this paper, we model sensor data calculation process and describe how to add data processing components onto content-based networks. On the contrary, data processing components causes concentrations of sensor data. We also describe load distribution mechanism of these components. Performance evaluations of our implementation shows that load distribution mechanism works well, and proposed system secure the scalability by adding data components.