Published papers

With a view to constructing a metropolitan network that can flexibly handle highly bursty traffic, an optical burst switched ring network with upstream prioritized switching and distributed fairness control is proposed. In optical burst switching, high burst drop rates at intermediate nodes constitute a problem. In the proposed system, burst signal dropping at the relay nodes is eliminated by means of upstream prioritized switching. At the same time, by using distributed fairness control, the problem of upstream prioritized switching-related unfairness, in which positional priority provides higher priority to more distant transmission nodes, is resolved. Further, by modifying the reservation of downstream nodes by using the trailer packet after disconnection of the transmission, unnecessary bandwidth is eliminated and the throughput is improved. By computer simulation, the throughput and the fairness with regard to the throughput are evaluated. It is shown that the proposed architecture exhibits excellent characteristics even in cases of hot-spot traffic, in which concentrated traffic occurs at a particular destination. © 2007 Wiley Periodicals, Inc.

A smart photonic cloud is defined as a universal network platform without any physical or logical constraint that provides flexible connectivity for machine-to-machine communication such as networked high-performance computing or intra- and inter-data center networks. The key requirements for the network in the Big Data era include an ultralarge capacity with low power consumption, low latency, as well as flexibility on demand to the changes in the configuration and bandwidth of the optical path. To cope with the growing demand for network virtualization, novel photonic layer virtualization will be proposed,which differs fromthe conventional approach in terms of the number of slices and the dynamic range of the bandwidth of each slice. First, the objectives and the guiding principle of the vision will be addressed. Next, three “Ss” will be presented that represent the key enabling technologies, namely scale-free photonics, smart photonic networking, and a synthetic transport platform.

Novel optical network architectures are proposed for creating future network services. The first architecture is a centralized approach for higher energy efficiency; it yields a data center-centric metro/access optical aggregation network based on wavelength/time-slot multiplexing. Not only higher application layer functions but also all layer-3 or upper traffic are transferred through the simple metro/access optical aggregation network and switched in a huge centralized giant router at the data center. Its simple optical switching is 200 times more energy efficient and only one electrical router is needed, so power consumption of the network can be reduced ten or twenty fold compared to the existing Internet. The second is service mash-up by linked data through a network that uses broadband optical wire for the IoT era. All service contents, hardware, and software programs are defined as service parts. Optical wire interconnects some service parts and creates new mash-up services in the network. That creates deep network functionality in combination with network and processing functions.

Elastic Lambda Aggregation Network (EλAN) has been proposed to improve the transmission efficiency and reduce the total energy consumption by integrating access networks of different services. In EλAN, Logical Optical Line Terminals (L-OLT) can be migrated between programmable OLTs (P-OLT). In this paper, the method of L-OLT migration for energy saving is proposed. L-OLT placement optimization problem is introduced to obtain the new placement that minimizes the number of running P-OLTs, and the migration sequence. P-OLT sleeping rate and the number of migrations are evaluated by simulation. The demonstration of L-OLT migration on the EλAN prototype system is also reported.

With the increase of traffic and energy consumption of network, energy-saving is an urgent matter. The realization of Multi-point to Multi-point service which can use network resources effectively is also needed. In this paper, we propose an SPMST (SuperPosed Minimum Spanning Tree) algorithm which can reduce the energy consumption of network by aggregating traffic on specied links and putting unused links into sleep mode. This can cover requirements of MP2MP MiDORi network. By computational simulation, the energy efficiency, short calculation time and capability of SPMST are presented. We veried practical effectiveness of proposed algorithm by using a prototype system of MiDORi network.

Machine-to-Machine (M2M) communication is expected to grow in networks of the future, where massive numbers of low cost, low function M2M terminals communicate in many-to-many manner in an extremely mobile and dynamic environment. We propose a network architecture called Data-centric Network (DCN) where communication is done using a data identifier (ID) and the dynamic data registered by mobile terminals can be retrieved by specifying the data ID. DCN mitigates the problems of prior arts, which are large size of routing table and transaction load of name resolution service. DCN introduces concept of route attraction and aggregation in which the related routes are attracted to an aggregation point and aggregated to reduce routing table size, and route optimization in which optimized routes are established routes to reduce access transaction load to the aggregation points. These allow the proposed architecture to deal with ever increasing number of data and terminals with frequent mobility and changes in data.

近年、サーバ、ストレージ、端末、ネットワーク等の技術進展により、クラウドサービスが急激に普及し始めている。また、サービスの多様化やユーザの増加により、サービス提供システムの拡大とともに消費電力が急増しており、システムの省電力化が重要な課題の1つとなっている。本論文では、複数のデータセンタとそれらを接続するネットワークより構成される大規模なクラウドシステムおよび提供するサービスに対して、サービスの品質を維持したシステムの省電力化手法について提案する。また、3都県に跨る大規模システムによる実証実験により、サービス品質を維持したクラウドシステムの省電力化について有効性を検証した。

Open Shortest Path First (OSPF) is one of the most widespread routing protocols in the world. In OSPF networks, routers calculate the paths for every traffic demand, based on weight values that are configured in advance. OSPF Weight Setting (OSPF-WS) is an NP-hard search problem; find the set of weights that maximizes network utilization. Internet Service Providers (ISPs) are facing the challenge of solving OSPF- WS within practical time and find the best weight set for the efficient use of the network. One heuristic approach, a scheme based on the Genetic Algorithm (GA), has been reported to offer fast solution of OSPF-WS. This scheme identifies good solutions comparable with the output of the conventional Integer Linear Programming (ILP) scheme. However, its calculation cost is still excessive for larger networks, thus higher processing performance is required. Unfortunately, the processing speed of single processing cores has become saturated, and the recent trend is a shift to multi-core processors. To best utilize the performance offered by these processors, the algorithm should be redesigned and suitably parallelized for multi-core CPUs. This paper redesigns the scheme of OSPF-WS with GA to create a parallelized algorithm with much lower computation overhead. Its performance is evaluated on a 16-core Intel Xeon processor and the result is a roughly 13 fold faster calculation speed than the original algorithm on a single-core CPU. This result shows the potential of further speedups with larger scale parallel processing units such as the GPGPU.

An extremely energy efficient layer-3 network architecture based on optical slot-type aggregation is proposed for the future Internet. Huge numbers of user traffic streams are multiplexed by optical wavelength/time-slots at the central data center router. Each time-slot is transparent, so all types of service can be integrated on any slot. In other words, user IP packets are aggregated and transferred through the Optical Aggregation Network to the centralized router, transparently. The proposed network architecture realizes a network structure well suited to traffic centralization, and reduces the power consumption to 1/10-1/20 compared to the existing Internet.

We report the first successful experiment on Point-to-Multipoint (P2MP) VLAN path establishment on the overlay-model-based GMPLS-controlled wide area Ethernet. To support the overlay model, P2MP VLAN path signaling with egress output port indication is proposed and implemented. It is confirmed that our extended RSVP-TE software can correctly establish P2MP VLAN paths in the overlay-model network.

Recently, attention is particularly focused on the research of Software defined network (SDN) for reducing network management complexity. The one of a key technology of SDN is OpenFlow. OpenFlow provide a centralized controller for network and the scalability of controller is main issue. In this paper, we propose a high-speed routing engine for improve the scalability of OpenFlow controller. Unlike conventional architectures of routing engine, the proposal is a hardware routing engine that using on-chip diorama network. We define the Diorama Network as a virtual emulated network in a chip. We implement a prototype of the routing engine on an actual dynamically reconfigurable processor (DRP), and test results show that the prototype can execute the shortest path calculation 19 times faster than the current approach.