We report on the demonstration of a fast restorable all-optical WDM network. This network consisted of four 44 optical cross-connects (OXC's) and four in-line optical amplifiers. These OXC's monitored not only the status of various network elements and quality of optical signals but also the optical path of each channel continuously. Thus, this network could automatically identify the causes of most network failures. For the fast restoration, we implemented these OXC's by using thermo-optic polymer switches (switching time: < 1.5 ms) and used hardware interrupt when LOS was detected. In addition, we used a pre-planned routing table made by using a simple heuristic routing and wavelength assignment algorithm. The results show that this network could be restored from any single link failure within 6 ms even when the restoration path was 400 km.

We demonstrate, through numerical simulations, the possibility of trans-oceanic single channel transmission at 160 Gbit/s with no active control. This was achieved using short period dispersion management, which supports short pulse propagation at practical map strengths. We demonstrate that through careful selection and optimisation of the system parameters the performance of this system can be extended. We also define the tolerable limits of the system to the residual dispersion slope and polarisation mode dispersion.

We investigated the characteristics of optical duobinary signals in achieving high fiber input power transmission focusing on the idea of optimum residual dispersion equalization. We confirm through calculations and experiments that setting the total link dispersion at a non-zero value allows high fiber launched power (+18 dBm) and large dispersion tolerance (350 ps/nm) at 10 Gbit/s. We demonstrate repeaterless 250-km single mode fiber (SMF) transmission with a 10-Gbit/s optical duobinary signal. We also demonstrate high-speed complete optical duobinary coding and transmit synchronous digital hierarchy (SDH) frames over optical duobinary signals for the first time.

A photonic crystal fibre has an array of microscopic air holes running along its length. The periodicity of the array is broken by a deliberate "defect" that acts as a waveguide core. Light is confined to this core by the holes. Although some designs of photonic crystal fibre guide light by total internal reflection and so can be considered analogues of conventional optical fibres, their properties can be strikingly different. Other designs guide light by photonic bandgap confinement and represent a totally new type of fibre.

The effectiveness and possible applications of all-optical signal processing using highly-nonlinear dispersion-shifted fibers (HNL-DSFs) are described. Transparent and simultaneous processings of multi-channels WDM signal are key features of optical fiber processors. Simultaneous wavelength conversion of 3210 Gb/s WDM signal by four-wave mixing, all-optical 3R regeneration of 220 Gb/s WDM signal using nonlinear loop mirrors, and simultaneous recovery of 2020 GHz WDM optical clocks by supercontinuum were successfully demonstrated using HNL-DSFs, and possible applications of ultra-fast and multi-channel processing in future photonic networks are discussed.

This paper describes a simple polarization-independent wavelength conversion method using degenerated four-wave mixing (FWM) in single-mode fibers pumped with cross-polarized high frequency, saw-tooth pulses from a single pump source. Successful polarization-independent wavelength conversion is experimentally confirmed with less than 12% and 5.6% variation using a gain-switched LD pumping and a mode-locked fiber laser pumping, respectively. We clarify that the interference effect between two orthogonal pump pulses must be taken into account to achieve a good polarization-insensitive operation, since even the small pulse edges bring about the large polarization fluctuations when they are interfered. Furthermore, it is reveal that the shorter pump pulse broadens its own spectrum due to the self-phase modulation in fibers, resulting in poor FWM efficiency. Finally, possibility of high-speed operation is discussed taking into account the pump pulse conditions.

We fabricated a 1.55-µm polarization insensitive Michelson interferometric wavelength converter (MI-WC). The MI-WC consists of a two-channel spot-size converter integrated semiconductor optical amplifier (SS-SOA) on a planar lightwave circuit (PLC) platform. Clear eye opening and no power penalty in the back-to-back condition were obtained at 10 Gb/s modulation. We also confirmed the polarization insensitive operation on the input signal. Moreover, for an application of the MI-WC to DWDM networks, we demonstrated the selective wavelength conversion of 2.5 G/s optical packets from Fabry-Perot laser diode (FP-LD) light to four ITU-T grid wavelengths. We confirmed the good feasibility of this technique for use in DWDM networks. The wavelength conversion we describe here is indispensable for future all-optical networks, in which optical signal sources without wavelength control will be used at user-end terminals.

We propose a novel optical add/drop multiplexer (OADM) utilizing free spectral range (FSR) periodicity of an arrayed-waveguide multiplexer (AWG). In this OADM, wavelength-division multiplex (WDM) signal is multiplexed and/or de-multiplexed in two steps. Power penalty due to coherent crosstalk is drastically reduced compared with that of conventional OADM where AWG multiplexers are opposite to each other. The calculated power penalty due to the coherent crosstalk is about 0.7 dB after the 16 OADMs in the case of 128 wavelengths. It was confirmed through a computer simulation that more than one hundred channels at 10 Gbps data rate could be accommodated in an OADM network with 16 nodes. These results show that the OADM network with over 1 Tbps capacity and 16 nodes could be constructed.

Recent studies on traffic measurement analysis in the various networks (LAN, MAN, WAN) have shown that packet traffic exhibits Self-Similarity. The packet traffic represents some behavior quite different from what it has been assumed. Some papers reported that Self-Similarity degrades the network performance, such as buffer overflow and network congestion. Thus, we need new network control scheme considering Self-Similar properties. The control scheme requires high-speed calculation method of Hurst Parameter. In this paper, we propose high-speed calculation method of Hurst Parameter based on the Variance-Time Plot method, and show its performance. Furthermore, we try to apply this method to the simple network control, in order to show effectiveness of the network control with Self-Similarity.

In the filed of cognitive psychology, simple recurrent networks are used for modeling the natural language processing in the human brain. For example, Elman experimentally showed that the simple recurrent networks can predict the right-most word in sentential forms of a particular grammar which can generate compound sentences with high probability. Concerning these results, it is natural to ask whether the computational capability of the simple recurrent networks is sufficient to recognize natural languages. In this paper, we assume that the range of a function computed at each gate of a simple recurrent network is a finite set. This is a quite realistic assumption, because we cannot physically implement a gate whose range is an infinite set. Then, we define equivalence relations between simple recurrent networks and Mealy machines or Moore machines, which are finite automata with output. Then, under our assumption, we show (1) a construction of a Mealy machine which simulates a given simple recurrent network, and (2) a construction of a simple recurrent network which simulates a given Moore machine. From these two constructions, we can conclude that the computational capability of the simple recurrent networks is equal to that of finite automata with output under our assumption.

Using the chirped grating with temperature control, we demonstrated the adaptive dispersion compensation at 40 Gbit/s RZ transmission. The simple monitoring of the 40 GHz frequency component enables us to automatic control of the adaptive dispersion compensator.

Multi-level divide-and-conquer (MDC) is a generalized divide-and-conquer technique, which consists of more than one division step organized hierarchically. In this paper, we investigate the paradigm of the MDC and show that it is an efficient technique for designing parallel algorithms. The following parallel algorithms are used for studying the MDC: finding the convex hull of discs, finding the upper envelope of line segments, finding the farthest neighbors of a convex polygon and finding all the row maxima of a totally monotone matrix. The third and the fourth algorithms are newly presented. Our discussion is based on the EREW PRAM, but the methods discussed here can be applied to any parallel computation models.

This paper conducts performance evaluation and performs simulation for a code division multiple access (CDMA) system when channel bands of multiple neighboring CDMA/DSSS are overlapped in time domain. It is assumed that all systems adopt direct-sequence spread-spectrum (DSSS) technique and are BPSK modulated by the different carrier frequencies. Automatic power control (APC) is also applied in the interfered system such that the receiver gets the same power from all users. Without loss generality, an additive white Gaussian noise (AWGN) channel is also assumed during analysis. In this paper, the analytic solution of the signal to noise ratio (SNR) is first derived in which both CDMA systems are modulated by different carrier frequencies. We have the results by simulation with Δ f = 0 and Δ f = 1 MHz, respectively. This analysis is good for general cases; and the results show an excellent computational performance. In particular, the result is very close to Pursley's result, when the systems have the same code length with no carrier difference.

A new algorithm called core-stateless random-early-detection (CS-RED) is described. This algorithm can improve the fairness characteristics under a wide range of network conditions and is robust in terms of the setting of its control parameters. It can be used in a distributed architecture, as can the core-stateless fair-queuing (CSFQ) algorithm, so it is applicable to large-capacity backbone routers. Through a detailed evaluation, we have identified and clarified the fairness degradation problem in the CSFQ algorithm. We solved this problem by incorporating the congestion notification proposed in the RED algorithm into CSFQ; we call this hybrid CS-RED. Comparison of the fairness characteristics in a computer-simulated network using either the CS-RED or CSFQ algorithm showed that a higher level of fairness is achieved with the CS-RED algorithm under a wide variety of network conditions without degrading link utilization. Furthermore, the CS-RED algorithm uses a treatment that is more general than the penalty box mechanism to penalize unresponsive flows in best-effort networks.

We improve the algorithm to obtain the min-cut graph of a hyper-graph and show an application to the sub-network extraction problem. The min-cut graph is a directed acyclic graph whose directed cuts correspond one-to-one to the min-cuts of the hyper-graph. While the known approach trades the exactness of the min-cut graph for some speed improvement, our proposed algorithm gives an exact one without substantial computation overhead. By using the exact min-cut graph, an exhaustive algorithm finds an optimal sub-circuit that is extracted by a min-cut from the circuit. By experiments with the industrial data, the proposing method showed a performance enough for practical use.

Adaptation of software components to the requirements is one of the key concerns in Component Based Software Development (CBSD). In this paper, we propose a formal approach to compose component based systems which are adaptable to the requirements. We focus on the functional aspects of software components and requirements, which are expressed in S-sorted functions. Those S-sorted functions are transformed into Colored Petri Nets (CPN) models in order to evaluate connectivity between the components, and to evaluate adaptability of composed systems to the requirements. The connectivity is measured based on colors or data types in CPN, while the adaptability is measured based on functional equivalency. We introduce simple glue codes to connect the components each other. The paper focuses on business applications, however the proposed approach can be applied to any other domains as far as the functional adaptability is concerned.

In this paper, we consider the net assignment problem in the logic emulation system. This problem is also known as the board-level-routing problem. There are field programmable logic arrays (FPGAs) and crossbars on an emulator board. Each FPGA is connected to each crossbar. Connection requests between FPGAs are called nets, and FPGAs are interconnected through crossbars. We are required to assign each net to the suitable crossbar. This problem is known to be NP-complete in general. A polynomial time algorithm is known for a certain restricted case, in which we treat only 2-terminal nets. In this paper we propose a new polynomial time algorithm for this case.

Recent progress in ultrahigh bit rate optical transmission using time-division multiplexing and wavelength-division multiplexing is described. Latest over 1 Tbit/s transmission experiments are shown and the possibility of higher bit rate transmission is discussed.

Method invocation mechanism is one of the essential features in object-oriented programming languages. This mechanism contributes to data encapsulation and code reuse, but there is a risk of runtime type errors. In the case of object-oriented databases (OODBs), a runtime error causes rollback. Therefore, it is desirable to ensure that a given OODB schema is consistent, i.e., no runtime error occurs during the execution of queries under any database instance of the OODB schema. This paper discusses the computational complexity of the type-consistency problem. As a model of OODB schemas, we adopt update schemas introduced by Hull et al., which have all of the basic features of OODBs such as class hierarchy, inheritance, complex objects, and so on. The type-consistency problem for update schemas is known to be undecidable. We introduce a subclass of update schemas, called acyclic schemas, and show that the type-consistency problem for acyclic schemas is in coNEXPTIME. Furthermore, we show that the problem for recursion-free acyclic schemas is coNEXPTIME-hard and the problem for retrieval acyclic schemas is PSPACE-complete.

This paper describes simple and efficient (linear-preserving) reductions between the Decision Diffie-Hellman problem and related problems.