グエン キエン

© 2019 IEEE. The dawn of 5G is rising with emerging wireless and network technologies that promisingly transform various aspects of human lives. The 5G mobile devices are going to tightly engage with the 5G infrastructure for future applications such as virtual reality, AI-enabled applications, etc. However, we argue that the tight engagement with convenience may bring downside that is a user may not correctly control/own her/his device. In particular, the user may not have the flexibility in utilizing networks, as well as, be aware of the network selection. This paper introduces a novel networking stack for 5G mobile devices (namely 5GVir) that leverages virtualization techniques for flexibility and awareness. The 5GVir device can concurrently exploit surrounding wireless networks by using Software Defined Networking (SDN). SDN provides a rich set of flexibility feature that will meet user expectation. Moreover, 5GVir includes wireless virtualization that can relax the dependence on hardware as well as provides a wireless link for the SDN's control channel. Last but not least, the 5GVir device has an additional awareness feature aware that drives each application process to predetermined networks (i.e., applying network namespace). Our prototype shows the potential of realizing 5GVir. Moreover, the initial experiments show that the virtualization in 5GVir has negligible overhead.

We introduce WiPoMu a cross-layer system aiming to improve performance and resilience of the traditional Wi-Fi model. Specifically, we integrate three technologies Wi-Fi virtualization, Policy Routing, and Multi- path TCP and let them work on a Wi-Fi client with commodity hardware.

This paper proposes a practical approach to on-demand access networks for Internet connectivity. Here, the access network is quickly established by commodity mobile devices which are carried by the users. After joining the network as a common station (STA), the device is transformed itself into a virtual access point (VAP), providing Internet connectivity to its vicinity. A network auto configuration software was developed and downloaded to the connected devices automatically for this VAP transformation. Implementation details of the wireless virtualization and VAP mechanisms which realize multihop Internet connected access networks are thoroughly presented. A preliminary prototype has been developed and evaluated via large-scale field experiments to confirm the feasibility and effectiveness of the proposed approach.This paper proposes a practical approach to on-demand access networks for Internet connectivity. Here, the access network is quickly established by commodity mobile devices which are carried by the users. After joining the network as a common station (STA), the device is transformed itself into a virtual access point (VAP), providing Internet connectivity to its vicinity. A network auto configuration software was developed and downloaded to the connected devices automatically for this VAP transformation. Implementation details of the wireless virtualization and VAP mechanisms which realize multihop Internet connected access networks are thoroughly presented. A preliminary prototype has been developed and evaluated via large-scale field experiments to confirm the feasibility and effectiveness of the proposed approach.

This paper proposes a practical approach to on-demand access networks for Internet connectivity. Here, the access network is quickly established by commodity mobile devices which are carried by the users. After joining the network as a common station (STA), the device is transformed itself into a virtual access point (VAP), providing Internet connectivity to its vicinity. A network auto configuration software was developed and downloaded to the connected devices automatically for this VAP transformation. Implementation details of the wireless virtualization and VAP mechanisms which realize multihop Internet connected access networks are thoroughly presented. A preliminary prototype has been developed and evaluated via large-scale field experiments to confirm the feasibility and effectiveness of the proposed approach.

We introduce a new way of handling fast failover on an OpenFlow switch, which is a fundamental component in Software Defined Networking (SDN). The fast failover is introduced in the scope of newest version of OpenFlow specification. Moreover, we present a new method for measuring the fast failover time and the results on an OpenFlow switch.

This paper proposes a practical approach for implementing disaster recovery access networks using on-site commodity mobile devices. In this approach, each device can work in two modes, namely station (STA) and access point (AP) modes, concurrently providing an effective way for multihop communications. Implementation details of wireless virtualization and virtual access point mechanisms realizing multihop Internet connected access networks for disaster recovery are thoroughly presented.

The Link Aggregation (LA) automatically distributes and load balances the traffic across the links within a Link Aggregation Group (LAG). Therefore, LA can improve network performances in terms of network resilience and throughput. On the other hand, the earlier versions of OpenFlow Switch Specifications did not provide LA because their forwarding model simply supported drop, forwarding, and flood packets. Fortunately, since the version 1.1 of OpenFlow Switch specification, LA has been added via the concept of virtual ports, hence the OpenFlow switches are potentially beneficial from LA. However, there is a lack of performance evaluation of LA in the OpenFlow context. In this paper, we investigate the performance of LA in a real testbed with two different OpenFlow platforms. The first platform is an open source Open vSwitch version 1.10.0 and the second one is an implementation based on the specification of OpenFlow 1.3. Both platforms are supported and running on top of the hardware Pica8 P-3295 Switch. The evaluation results show that both platforms can balance the network load on each link of a LAG. Consequently, the aggregation bandwidth is increased. Moreover, the OpenFlow 1.3 based platform provides a better resilience in comparison to the Open vSwitch platform. Specifically, the OpenFlow 1.3 based platform spends less than 1.508 seconds to switch the TCP traffic on the faulty link over the other links of a LAG while this time is four times longer in the other platform. Similarity, in the case of UDP traffic, the maximum switchover time on the the Open vSwitch platform is twice the longest switchover time on the OpenFlow 1.3 based platform.

Linux Ethernet bonding is a feature that allows grouping multiple physical network interfaces into a logical one, aiming to improve fault tolerance and network throughput. Although Linux bonding has been deployed in various network scenarios, its performance has not been well investigated. Addressing the problem, in this paper we present an evaluation of Linux bonding in order to explore its capabilities in a real testbed. Different from other works, in the evaluation we consider the traffic flows of network applications as well as examine various modes supported by Linux bonding. The evaluation results show that Linux bonding can provide the flow switch-over time - the duration of traffic flow discontinuation due to a network failure - as small as 10 milliseconds. Moreover, in a non-failure scenario with file transfer protocol (FTP), the Linux machines with bonding reduce the time of file transfer process to nearly a half comparing to the one in case of without bonding.

Disconnection of Internet access in case of natural disasters may cause serious obstacles for relief works. Immediately providing Internet connection in disaster regions is still a challenge because infrastructure-based networks may have been disabled by the disaster. This paper proposes a Group-based Internet Connection Spreading (GICS) architecture for disaster recovery. This architecture supports groups of unconnected mobile devices to connect to the Internet through still alive infrastructure access points or groups of Internet-connected mobile devices in an effective way. By the support of the Internet-connected Device Discovery mechanism in proposed GICS, mobile devices can discover nearby devices that have already reached the Internet connection. GICS enables mobile devices within groups collaborate together to bring reasonable Internet connection to all group members. Moreover, GICS can flexibly adapt with changes in network condition since GICS groups can necessarily rearrange the roles of group members to maintain the suitable the Internet connection. In addition, the quality of connection between devices in GICS is reliable based on IEEE 802.11 (Wi-Fi) standard.

It is highly required for an effective way for quickly setting up networks and keeps Internet connectivity for victims in disaster affected areas. The network must be set up using on-site commodity mobile devices which are commonly used in daily life. More concretely, no special equipment or additional hardware is required for network establishment. Moreover, the network must be automatically set up or at least it must be easily to be configured by ordinary people without any technical skill. These serious conditions in disasters make the task hard to be solved. In this work we propose a new approach by which wireless virtualization and software based access point technologies were well utilized for quickly setting up the access network to provide Internet access. The real field experimental results reveal the feasibility as well as the usability of the proposed approach.

地震や津波などの災害はすべてを破壊し、通信も途絶して人々を孤立させてしまいます。このため、情報通信ネットワークは災害からすぐに復旧して再構築されるようにすることが重要です。この研究では、被災地で人手を要せずに、すぐに構築できる災害復旧用アクセスネットワーク技術を提案しています。具体的には、情報通信ネットワークの一部が破壊されても、被災地にあるモバイル端末 (スマートフォン、ラップトップ PC、タブレット端末等) を、生き残っているネットワークの端までマルチホップで接続することによって、被災地の人々が被災直後からインターネットを自由に使えるようにすることを目指しています。

現代社会では、ICTが必要不可欠なものとなっており、大規模な災害に見舞われた場合でも、ICTの継続的利用が求められている。一般企業でも、いわゆるBCP(事業継続計画)の一環で、社内ICTのレジリエント化が期待されている。本稿では、複数のサーバ、端末をネットワーク化した一般的なICTについて、そのレジリエンス性を向上させるサーバや網の仮想化等の各種技術の現状調査をもとに、レジリエントなICTに求められる要求条件を明らかにする.

We propose a new synchronous MAC protocol with QoS awareness named AQ-MAC. AQ-MAC adopts a receiver initiated manner and provides QoS service per packet following the priority imposed. When a node has an incoming packet with high priority,it immediately turns on the radio in order to wait for a transmission initiated by a receiver. Otherwise the node keeps the low priority packets in a queue and sends out in a burst until a high priority packet comes or after a timeout value. In addition, AQ-MAC utilizes a packet concatenation scheme to improve the energy efficiency by reducing control overhead. We have evaluated AQ-MAC in multiple scenarios using ns-2. The results show that AQ-MAC adapts well with different types of traffic as well as achieves good performances of energy efficiency and latency.

In this paper we investigate the impact of contention on the flow capacity of Demand Wakeup MAC (DW-MAC), which is a state-of-the-art multi-hop MAC protocol. We find that instead of using the original large contention window (CW), a DW-MAC's flow using a smaller CW could avoid unnecessary contention. The analysis and ns-2 simulation result revealed that DW-MAC with new values for CW achieves higher throughput, lower end-to-end latency, and greater energy efficiency.

In wireless sensor networks (WSNs), MAC protocols generally employ a duty cycling mechanism according to which the sensors frequently turn on and off their radio module in order to conserve battery power that would otherwise be wasted by energy-costly idle listening. In duty-cycling MAC protocol, packets often wait in nodes' queue since the sleep duration of radio is much longer than the active duration. Moreover the packets in sensor network usually are forwarded via multiple hops to the sink. These characteristics motivate the utilization of packet aggregation. Using packet aggregation scheme not only increases throughput but also reduces power consumption that is incurred by additional control packets. Since data packets in sensor networks are comparatively small significant gains are to be expected. In this paper we proposed the packet aggregation scheme (PA) which can be applied to the duty cycling MAC protocol in WSN. We evaluated PA with both single hop MAC and multi-hop MAC protocols using the network simulator ns2. The results show that the new protocols combined with PA outperform the originals both in terms of energy efficiency and throughput.

In this paper we propose a new method to improve energy efficiency of the multi-hop MAC protocol in wireless sensor networks. This method uses carrier sensing technique to adapt to the traffic status of the network. By using carrier sense, nodes will decide whether to sleep or to keep awake depending on the existence of data. Simulation results showed that using new method can achieve higher energy efficiency while slightly increase latency.

In this paper we propose a new duty cycling medium access control (MAC) protocol with low latency and low control overhead for wireless sensor networks (WSNs) use (the LCO-MAC). In LCO-MAC, a DATA packet can be transmitted through multiple hops in a single duty cycle to shorten end-to-end latency. Exploiting the broadcast property of wireless network, we employ one packet to play more than one role to reduce energy consumption caused by control packet overhead. Our simulation using ns-2 has shown that LCO-MAC enables a notable improvement in energy efficiency and decreases end-to-end latency compared to those of RMAC.