Hiroshi Masuyama

ON FAULT TOLERANCE OF HYPERCUBES USING SUBCUBES

A hypercube is an important interconnection topology in parallel and distributed processing. This paper treats fault tolerance of a hypercube. More precisely, it discusses relations between a faulty hypercube and its fault-free subcubes. First, this paper presents an upper bound on the minimum number of faults in an n-cube where no fault-free (n - m)-subcube can exist. Next, it is shown that the bound can be improved if m = 2. Finally, the paper discusses the number of faults when there are always at least two fault-free disjoint (n - 2)-subcubes.

New Optimal Channel Assignment for Hierarchical Cellular Networks

This paper discusses the channel assignments which lead to the highest possible communication connectivity on the condition that the total number of required channels is constant. Several channel assignment schemes have been proposed depending on repacking approaches; non-repacking, always repacking, and repacking on demand. In this paper, these schemes are evaluated primarily with the prospect of attaining the highest communication connectivity. The evaluations quantitatively indicate that repacking on demand may significantly outperform the others under the condition that not only the number of prepared micro-cell channels remains constant, but also the total number prepared of both micro-cell and macro-cell channels remains constant as well

ALGORITHMS TO REALIZE AN ARBITRARY BPC PERMUTATION IN CHORDAL RING NETWORKS WITH FAILURES

A family of regular graphs of degree 3, called chordal rings is presented as a possible candidate for the implementation of a distributed system and for fault-tolerant architectures. The symmetry of graphs makes it possible to determine message routing by using a simple distributed algorithm. Arbitrary data permutations are generally accomplished by sorting. For certain classes of permutations, however, there exist algorithms that are more efficient than the best sorting algorithm. One such class is the Bit Permute Complement (BPC) class of permutations. In this paper, we first develop algorithms requiring two token storage registers in each node to realize an arbitrary BPC permutation. We next evaluate its ability to realize BPC permutations in networks of arbitrary size by estimating the number of required routing steps when a single fault is present and when not.

Hierarchical System-Level Diagnosis Based on a Voting Scheme for Identical-Unit-Interconnection Systems

A hierarchical non-adaptive system-level diagnosis approach is introduced categorized into three types of interconnection systems; array-, hypercube-, and tree-typed systems which consist of a large number of identical units (This type of interconnection system does appear to be more common in modern hardware such as red storm supercomputer or blue gene/L system in parallel computers). This diagnostic approach employs a simple voting scheme where the votes are counted locally by a monitor unit in each set and in each hierarchical cluster, and the monitor calculates an answer based on the threshold. The main contribution of this paper is to show the effectiveness of this test architecture under realistic circumstances.

New non-adaptive distributed system-level diagnosis methods for computer networks

Abstract: A hierarchical non-adaptive diagnosis algorithm is presented for testing total N nodes of computer networks. Since general computer networks can be regarded as an N -nodes complete graph, then for the efficient testing, it is essential that the test process be parallelized to enable simultaneous test of multiple nodes. In order to attain this object, we propose a noble test graph enabling to test as many nodes as possible in a network due to a hierarchical architecture of test processes. The amount of test times is evaluated as the diagnosis latency. Optimal diagnosability t is analyzed under clustered fault distribution. In order to reduce the amount of required test times, two revised approaches are discussed and evaluated.

All-to-All Personalized Communication Algorithms in Chordal Ring Networks

Chordal ring networks are ring structured and they become unreliable when nodes or arcs in the ring break down. Such a shortcoming can be circumvented by adding duplicate rings. Though this inherent topology of preparing many loops in the network structure is advantageous to broadcasting, no such broadcasting algorithms have been reported up until now.This paper discusses broadcasting algorithms for Chordal ring networks under the last one of four different communication primitives: One-to-all broadcasting, one-to-all personalized communication, all-to-all broadcasting, all-to-all personalized communication. In this paper, Chordal ring networks are assumed to be not under critical faults, that is, the maximum number of faults we consider is 2. The number of time units required to the proposed algorithm which can tolerate under to the critical faults is made clear.

Energy Lesser Broadcasting Algorithms Using Adjustable Transmission Ranges in Mobile Ad Hoc Networks

This paper discussed 2 approaches to communication algorithms; 2-level clustering mesh approach and 1-level flat mesh approach, and proposes one of them which makes it appear that the total amount of expended energy becomes lesser. Wu and Dai [1] previously proposed 2 approaches; 2-level clustering approach and 1-level flat approach. This study [1] is based on the virtual backbone conception. Our 2 proposed approaches change the clustering performed in [1] into mesh so that energy consumption becomes smaller in a broadcasting. Total energy is required in not only the broadcasting process but also the construction process of a broadcasting tree. This is our approaches become disadvantageous. In [2], we gave the static properties of our approaches. This paper gives the advantageous bound of mobile movement for the 1 level flat mesh approach through our simulation study in realistic situations where a certain percentage of total mobiles move at random speeds.

A location management method for cellular networks

A location management method is discussed in reference to the classical model of cellular networks based on hexagonal cells. This location management forms two phases; location update and paging phases. The most important criterion of the best location management is to find the optimum interval of location update which results in the minimum energy consumption in a total process of location management. In this paper, the relationship between this optimum location update interval and the speed of the mobile terminal movement is investigated. Simulation results present the optimum location update interval for mobile speeds.

Algorithms for Energy-Efficient Broad- and Multi-Casting in Wireless Networks

This paper addresses the problem of broadcasting (and multicasting) focusing on the two points of energy efficient networking and of time efficient computing, where all base stations are fixed and each base station operates as an omni-directional antenna. We developed one broadcasting algorithm based on the Stingy method and based on the above performances. We evaluate this and the other two algorithms based on the Greedy and Dijkstra methods.

A Conference Network for Group Communications Based on Multistage Interconnection Networks

A conference refers to a group of members in a network: which communicate with each other within the group. In this paper, we study the construct of a conferencing server based on multistage interconnection networks which supports the simultaneous transmission of processed messages among conference participants. That is, we investigate the ability of multistage interconnection networks to support the conferences. Since multicasting and routing schemes in the networks have been developed, multistage interconnection networks can be used as a conferencing server. The key point to investigate is the trade off between the communication delay and the hardware cost. The superiority of this scheme is of no use to arbiters. The hardware cost required in this scheme is O(nlogn) for an ntimesn conference network. In order to evaluate the communication delay, performance evaluations are conducted in the networks which serve several kinds of groups, where n=16, 32, and 64. The performance results show that the total number of required cycles goes over n as the load factor is close at 1