Masahiro Tsunoyama

A fault-tolerant FFT processor

A scheme for concurrent fault detection by recomputing and a fault-tolerant FFT processor using the scheme are proposed. An FFT processor with perfect shuffle is considered. The realization of the processor is based on a linear cellular automaton (LCA) model having the constant-weight and equidistance properties. When a fault occurs in the processor, the fault is detected concurrently and the processor is reconfigured by replacing the faulty butterfly unit with a normal one according to the state of the processor. The reconfiguration can be made within a clock period by making a state transition based on the LCA model and by reconnecting the butterfly units according to the new state. The processor can be reconfigured quickly, so that it can be used for highly reliable real-time data processing systems.<<ETX>>

A fault-tolerant parallel processor modeled by a two-dimensional linear cellular automaton

A number of methods for constructing fault-tolerant parallel processors using one dimensional linear cellular automaton model have already been proposed. When the methods are used, however, a large amount of hardware is required to realize the connections among processing elements for executing a given algorithm, and for reconfiguring the processor quickly when faults occur in the processor. Moreover, these processors require long communication distances between processing elements so that the method is difficult to be used for large scale processors. In this paper, a method for constructing parallel processors based on a two dimensional cellular automaton model is proposed to alleviate the problems. In the processor based on this model, processing elements are connected in two dimension and the average communication distance can be reduced and a higher fault-tolerance due to the increase in the number of communication paths from 2 to 4.

A synchronization mechanism in networked virtual environment

We describe the design of a group synchronization mechanism for a local area network virtual environment and introduce a step synchronization mechanism. We use a football game to evaluate this mechanism. In this football game, users can share a virtual space where they can move around, dribble and shoot. To achieve synchronization among users, a client of the game exchanges data through a server within a certain period time. A problem of such a system is that all users should share the space in the same time for justice competition. But because of the different capacity of computers and the time cost of the communication between clients, it is difficult for all clients to update the virtual space simultaneously. In our systems, we introduce a waiting time synchronization mechanism. To evaluate our proposed architecture and mechanisms, we make a series of experiments in our local area network and then build and analyse the model of our system.

Approximation method for probability distribution functions using Cox distribution to evaluate multimedia systems

Several probability distribution functions such as exponential distribution function have been used to represent the task arrival process and processing time of tasks for multimedia systems. These distribution functions are simple, memory less, and easy-to-use, however,they are difficult to represent the task arrival process of multimedia stream data or audio data having real-time properties. This paper proposes a method for obtaining approximate probability distribution functions for given task arrival process for multimedia systems by using the Cox distribution function. The Cox distribution function can represent arbitrary distribution functions but their parameters are difficult to be determined for representing the given function. In the method, first, an objective probability distribution function is approximated by a linear combination of exponential distribution functions and Erlan distribution functions, then the parameters of the Cox distribution function is determined from the linear combination of the functions obtained above. The examples of the approximations show that the method is effective for approximating given probability distribution functions.

A concurrent fault-detection scheme for FFT processors

This paper proposes a concurrent fault-detection scheme for FFT processors. In the scheme, fault detection is made by comparing the pair of outputs from butterfly units based on the FFT algorithm. The hardware overhead for the scheme is O(N) where N is the number of input data. This scheme requires no extra computations for locating a pair of faulty butterfly units, therefore, the scheme can be used for highly reliable real-time systems.

An Application of Fuzzy Measure and Integral for Diagnosing Faults in Rotating Machines

This paper shows an application of fuzzy measure and fuzzy integral for diagnosing faults in rotating machines. The fuzzy degrees for fuzzy set of vibration spectra are determined by the membership functions for spectra. The functions are composed using values in the syndrome matrices given by the knowledge of skilled engineers. The fuzzy measures for the set of spectra and for the groups are obtained from the weights of spectra and weighs of groups based on the definition shown in this paper. Both of the weight s are shown in the syndrome matrices and given by skilled engineers.

An evaluation method for delay time and its jitter of WLAN using a GSPN model

This paper proposes a method for evaluating mean delay time and its standard deviation (jitter) of wireless local area networks (WLAN) using Generalized Stochastic Petri Nets (GSPN). The WLAN is modeled using GSPN and the mean delay time and its jitter are calculated using equations derived from the Markov chain associated with the GSPN model. The method is applied to evaluate delay time and its jitter for 802.11e WLAN. The results reveal that the values calculated based on the equations agree well with the values obtained from simulations.

A method for analyzing a queuing network by the tagged‐task model

This paper considers faults in distributed real-time systems modeled by queuing networks; in particular, a fast method of analysis of transient variation of system performance is proposed. This method involves analyzing queuing networks by means of the tagged-task model. Such an analysis requires repetitive determination of the limit-state probability at varied initial values; however, the computation time can be reduced by using the absorption probability (the probability of arriving at the absorption state). In addition, a fast algorithm is proposed to calculate absorption probability, in which the state set is split into cosets according to arrivability, and the cosets are replaced by representatives as soon as the state probability becomes low. Finally, the effectiveness and accuracy of the proposed method are verified by actual examples.