研究発表

A highly-energy-efficient network using an optical aggregation network and a service cloud has been proposed to reduce the power consumption of today’s Internet greatly. The optical aggregation network connects a solitary giant router to users by logical tree topology. This paper proposes a novel optical aggregation network architecture. The proposed architecture uses 2x2 optical switches and combines the features of a tree topology and a ring topology. The paper also introduces a design method of the proposed network architecture. It is shown that the method can design high-availability and low-optical-loss network with fewer optical switches compared to a duplex tree topology.

This paper focuses on parallel data transmission, which is one of the solutions in considering high-speed transmission in next generation wide area layer-2 network. To realize parallel data transmission, reliable delay and route measuring for every possible path between two nodes and choosing multiple paths with small delay difference is necessary. This paper proposes a new multipath discovery and delay measurement method using bi-directional flooding technique. In the proposed method, extended Ethernet frame and bi-directional flooding technique which runs on data plane is introduced. The computer simulation showed that by applying the proposed bi-directional flooding technique, scalability has been greatly increased.

This paper proposes high energy-efficient IP network architecture with Giant Center Router and Optical Aggregation Network. User IP packets are aggregated and transferred to Giant Center Router through Optical Aggregation Network transparently by dynamic optical resource allocation. Giant Center Router is power scalable with an amount of traffic proportionally. Optical Aggregation network is realized by circuit switches, wavelength converters, and wavelength multiplexers/ demultiplexers. The proposed network scheme reduces the power consumption to 1/20-1/30 compared to the conventional Internet, realizes a network structure well suited to traffic centralization, and offers smooth migration from the conventional Internet.

We have studied a novel optical access network with PLZT (Plomb Lanthanum Zirconate Titanate) optical switches,called ActiON (Active Optical Network), as a next generation access network to replace PON (Passive Optical Network). On the other hand, energy consumption of the present network becomes an important problem today. To resolve this problem, we study an energy efficient metro-access hybrid network by using ActiON. Also, it is necessary to extend a bandwidth capacity of ActiON to realize a scalable network. In this paper, we mainly focus on an energy consumption and a scalability. As a result, the proposed next generation metro-access hybrid scalable network accommodates 600000 users. In the evaluation, we show that proposed network reduces 65 % of energy consumption compare to the present network. Furthermore, blocking ratio of proposed network is lower than a quality assurance of the Internet service in any cases.

Recent years, cloud computing which provides computational resources via Internet is attracting the world. Electricity consumption of data centers increases rapidly because the number of users of data centers increases with the growth of cloud computing. According to the world wide trend which urges growing out of dependence on fossil fuel and nuclear energy, it is hoped to consume electricity effectively and wisely. Then, we propose Live Data Center Migration, which can use electricity wisely because functions of data centers are transferred to other data centers geographically apart. For this proposal, we expand the live migration technology into wide area. This technology is originally used inside a data center, and able to transfer virtual machines from the server to another without stopping processes of the virtual machines. Fig. 1 shows the basic concept of our proposal network. The networks among the data centers are set up over the NGL2 (New Generation Layer 2) network with GMPLS (Generalized Multi-Protocol Label Switching) before the virtual machines migration. During the live migration, the memory images of the virtual machines are transferred via this network. The routes between users and data centers have to be switched smoothly after the virtual machine migration. There are two methods to solve this problem. The first is that, packets arrived at DAT (Dynamic Address Translation) are encapsulated with the header addressed to the data center, and transferred via Internet. The route can be switched due to changing the destination address of capsulation header. The other is that, packets arrived at NGL2 switch are added VLAN (Virtual LAN) tags, and transferred via the NGL2 network. It is necessary to set up VLAN path between users and data centers in advance and the route can be switched due to changing VLAN tag. The proposed system is implemented as shown in Fig.2. In this presentation, we mainly discuss the protocol of how to set up the networks among data centers and how to switch the routes between users and data centers.

We propose a slot-switching based all-optical network to support various network services. Aggregating various types of networks (e.g., access networks for residential users, leased lines for business users, mobile backhauls, and fibre channels for data centers) in a single network is discussed as a candidate of next generation access network. Figure 1 shows the concept of this network. There are virtual OLTs (Optical Line Terminals) in a solitary programmable OLT and each of them provides different services at respective protocols. By integrating network equipments, this network reduces energy consumption drastically. It also facilitates the introduction of new network services which use their own protocols. The simplest way to realize this network is using WDM (Wavelength Division Multiplexing) and allocating each wavelength to each network service. However, this method has low scalability because the number of wavelengths that are able to multiplex in a single fiber is limited. In addition, CAPEX (CAPital EXpenditure) becomes high because it is necessary to develop various types of transceivers that adapt different wavelengths respectively. Our proposed network is an expansion of Active Optical access Network (ActiON) that we have proposed earlier, and uses high-speed optical switches. As shown in Figure 2, this network has capability to transfer any data signal formats transparently in a single wavelength. The bandwidth is shared by all users and dynamically allocated by fixed-length time slots. In this presentation, we mainly discuss the network architecture.

With the advance of the ubiquitous society by shift to IPv6, it is expected that various devices such as cars, home electronics, sensors, tags, and so on will be allocated IP address and connected to the network environment. On the other hand, Cloud computing is attracting the world. Cloud computing, via Internet, provides computation, software, data access, and storage services that do not require end-user knowledge. Then, we have been proposing IP based uGrid (ubiquitous Grid Networking Environment). In uGrid, everything connected to the network is defined as “Service-Part” and assigned IP address. Service-Parts mean not only devices but also software programs and data functions which are virtually “things”. Users can use Service-Parts all over the world. Moreover, by combining various Service-Parts via network, uGrid can provide new mash-up services. Figure 1 shows a concept of IP based uGrid. By connecting several IP addresses, the mash-up service of Service-Parts is realized. Because uGrid manages Service-Parts and executes uGrid service provisioning based on IP address, we utilize the IP mechanism such as routing, signaling, and DNS. The merit of utilizing IP mechanism is that, IP is an originally scalable mechanism, and many sophisticated technologies accumulated for years are available. As a technology to manage Service-Parts with IP addresses, GMPLS is applicable. GMPLS is a suite of protocols which allows automate the provisioning and management of the network. Originally, GMPLS was proposed to generalize MPLS for packet switched network to other networks with different switching capabilities like TDM, lambda, fiber, and so on. Our motivation is to extend this “generalized” from switching capability (network, data link, and physical layer) to Service-Parts (application layer). In this research, extended routing protocol (Service-Routing) and extended signaling protocol (Service-Signaling) are proposed. Service-Routing searches Service-Parts and calculates the route for service mash-up. Service-Signaling executes reservation of Service-Parts resources and the establishment of the connection among Service-Parts. By experiment, it is shown that the combination of Service-Routing and Service-Signaling was succeeded. This technology can realize next generation inter-Cloud which integrates the various Cloud such as SaaS, PaaS and HaaS/IaaS. By interconnecting various IP addresses of services via high-speed and high-capacity optical path, mash-up of service, as if executed in one server without considering the network distance, in cloud network can be realized.

近年, インターネットトラヒックの増加に伴い, インターネットにおける消費電力が増加しており, 消費電力の削減手法として光アグリゲーションネットワークが提案されている. 光アグリゲーションネットワークは, インターネットの多数のルータを1 台の大容量ルータに集約する大規模なネットワークである. したがって, 上り通信において動的帯域割当(DBA:Dynamic Bandwidth Allocation) を効率的に行う必要がある. 本稿では, 光アグリゲーションネットワークにおいて, スループットを向上するスケーラブルなDBA 方式を提案する. 提案方式ではONU をOLT (Optical Line Terminal) からの距離別にグルーピングすることでDBA 計算時間およびRTT 差によるアイドル時間を削減し, スループットを向上可能であることを示した.

我々は, イーサネットをベースとした新たなレイヤ２プロトコル技術(New Generation Layer-2: NGL2)を提案し, 要素技術の開発及びプロトタイプシステムでの実証実験を進めてきた. 今回, NGL2 over 既存広域イーサネット網実証実験の一環として, 100GbE対応のテラビットLANシステムとの連携実証実験を行った結果を報告する.

This paper reports the demonstration results of the dynamic link power management by energy optimal network topology design with prototype Gigabit layer-2 switches. The result shows dynamic link power management is feasible.

信頼性の高い広域レイヤ2ネットワークの実現に向けて，一時的なフレーム転送を用いた高速な障害回復機能が提案されている．従来では，障害対策用のフォワーディングテーブルに基づき障害検知ノードが一時的なフレーム転送を行っていた．しかし，本テーブルはパスのあて先を考慮しないため未到達によるフレームロスが発生する可能性があった．そこで本研究では，k-hopフラッディングを用いた一時的なフレーム転送方式を提案する．本方式では，障害検知ノードから全隣接ノードへの転送によりフレームロスが削減可能となる．計算機シミュレーションにより，未到達によるフレームロスおよび過剰帯域によるフレームロスの評価を行い本提案方式の有効範囲を示す．

我々は，PON (Passive Optical Network) に代わる次世代光アクセスネットワークとして，収容加入者数の増加と伝送距離の伸長を可能とした，PLZT (Plomb Lanthanum Zirconate Titanate) 光スイッチを用いる新たなアクティブ型光アクセスネットワークActiON (Active Optical Network) を提案している．本稿では，PLZT 超高速光波長選択スイッチを用いることで60 万ユーザの収容を可能にしたメトロ・アクセス融合型次世代スケーラブルネットワークを提案する．特性評価において，現在のネットワークにおいて問題とされている消費電力に対し，提案のネットワークでは消費電力を65%削減可能であることを示した．また，60 万ユーザを収容した提案ネットワークにおけるブロッキング率をインターネットサービスにおける保証値以下に抑制した．

現在のインターネットの消費電力を大幅に削減する手法として，1台の大容量ルータと数千から数万のユーザを1ホップで接続する光アグリゲーションネットワークが提案されている．本研究では，多段に接続した2x2PLZT光スイッチを用い，ツリー型トポロジとリング型トポロジの特徴を組み合わせることにより，低光損失および高信頼性を満足する全光アグリゲーションネットワークを構築する手法について検討を行った．

近年,インターネットの普及に伴いネットワークのトラヒック量及び消費電力が増加しており,ネットワークの省エネルギー化が重要な課題となっている.そこで我々は, TE(TrafBc Engineering)によりトラヒックを集約しネットワーク機器のポート単位で電源のON/OFFを行うことで省エネルギー化を目指すMiDORi(Multi-(layer,path, and resources) Dynamically Optimized Routing)ネットワークを提案している. MiDORiでは, NM(Network Manager)が各スイッチに設定を行っているが,大規模ネットワーク再構築の際に設定時間が膨大になるという問題がある・そこで本稿では,自律制御により高速設定が可能な手法として, GMPLS(Generalized Multi Protocol Label Switching)を用いたアーキテクチャを提案する. GMPLSではシグナリングによりパス上のポートを一括して電源制御が可能である.しかし,同一ポートを複数パスで共有しているシングルレイヤでのGMPLS適用の場合,パス切替時の電源制御により他のパスが断絶する問題が発生した.そこで,シグナリングの際に各ノードがネクストホップ情報を保持することで,他のパスに影響を与えないパス切替を可能とした.実装実験および計算機シミュレーションにより, 5ホップにおいて従来より79%パス設定時間を削減でき,また再構成においてパスの断絶が発生せず,提案方式の有効性を示した.

With the advance of ubiquitous society including IPv6, it is expected that everything will be connected to the networking environment. Then, a ubiquitous grid networking environment (uGrid) has been proposed. In uGrid, everything from a device to a program is defined as “Service-Part”, and new mash-up services can be provided by connecting several Service-Parts. In this paper, IP Routing/Signaling-based uGrid service provisioning is proposed as a new concept for realizing uGrid. The feature of this concept is to assign IP addresses to Service- Parts and execute uGrid service provisioning in the network layer (IP protocol layer). IP protocol is an originally scalable protocol. Also, many sophisticated technologies accumulated for years can be used in IP protocol. Therefore, by using IP protocol, the construction of scalable and intelligent uGrid is expected. As key techniques, Service Routing and Service Signaling are introduced, and the issues of each technique are listed. By experiment, it was confirmed that the provisioning of a simple video grid service was succeeded by Service Signaling.

我々は超高速光スイッチを用いたアクティブ型光アクセスネットワークActiON (Active Optical Network)の提案を行っている. ActiONは現在普及しているPON (Passive Optical Network)と比較して, 収容加入者数の増大および伝送距離の伸長を実現する. これにより, PONによるカバーが困難な山間部や中国等の人口密集地域への光アクセスネットワークの導入を図る. また近年では, 音声通話等のライフラインサービス, 業務取引等のミッションクリティカルな通信がネットワークを介して行われており, 次世代光アクセスネットワークにおいて信頼性の向上は必須である. 本発表ではActiONにおける高信頼化に着目し, 光スイッチを活用したプロテクション方式について提案を行う. ActiONでは光スイッチにより柔軟な冗長構成および予備系への高速切り替えが可能となり, 費用対効果の高いプロテクションが可能であることを示す.

クラウドネットワークの登場により, データセンタ間における高速な通信が必要となっている. しかし従来のインターネットを介した通信では, 帯域幅が不十分であるという問題点がある. そこで我々は, 安価な専用線を複数使用し並列伝送を行うことにより擬似的に広帯域なWANを構築するインテリジェントクラウドネットワークの提案を行っている. 本稿では, GMPLSを用いて並列経路のパス確立を行う並列伝送システムを提案する. 本システムをソフトウェアスイッチに実装し, End-to-Endの実効帯域測定による並列伝送の評価およびパス確立完了までの所要時間の計測を行う.