Showing papers by "Hyoung Joong Kim published in 1996"
TL;DR: An algorithm which optimally determines the order of load distribution is developed and it is shown that the order depends only on the channel capacity between nodes but not on the computing capability of each node.
Abstract: A tree network consisting of communicating processors is considered. The objective is to minimize the processing time by distributing the processing load to other nodes. The effect of the order of load distribution on the processing time is addressed. An algorithm which optimally determines the order of load distribution is developed. It is shown that the order depends only on the channel capacity between nodes but not on the computing capability of each node.
54 citations
TL;DR: Numerical simulations show that the proposed method for solving ill-conditioned linear equations is effective enough to be applied to the practical finite element problems.
Abstract: A new algorithm is presented for solving ill-conditioned linear equations. This method is radically different from conventional methods. We improve the condition of the linear equation first and then solve the well-conditioned one. The approximate solution is modified to get the exact solution. Numerical simulations show that the proposed method is effective enough to be applied to the practical finite element problems. Mathematical soundness of the algorithm is also shown. Several implementation techniques are discussed. It can be used not only for solving ill-conditioned equations but also for inverse problems and many other areas.
19 citations
27 May 1996
TL;DR: The simplest diagonal scaling is a good alternative for improved convergence and easy parallelization and data communication time can be reduced at the cost of convergence rate, MATLAB simulation result is given.
Abstract: Network-based computing with cluster of personal computers or workstations recently has become an important and successful technique. Network-based computing enables fast computation and resolves short of storage with computers on the desks. A cluster of inexpensive computers offers them aggregated computing power and storage to challenge very large-scale problems. Iterative solvers are used for the large sparse linear systems. Preconditioners accelerate the rate of convergence of the iterative solvers. However, parallelization of preconditioners is far from satisfactory so far. Thus, the simplest diagonal scaling is a good alternative for improved convergence and easy parallelization. In order to reduce communication overheads, a data compression technique is considered. Data communication time can be reduced at the cost of convergence rate, MATLAB simulation result is given in this paper. This method is effective to cluster of computers connected through slow networks.
3 citations
TL;DR: A modified Rayleigh quotient iteration (MRQI) that can always locate the minimum eigenpair is proposed and Mathematical properties of the MRQI are investigated.
1 citations
01 Jan 1996
TL;DR: Design principle of disk arrays for video-on-demand or multimedia- on-demand server, size of striping units, data request size, data placement strategies and other important issues related with playback servers are discussed.
Abstract: Basic requirements of video-on-demand or multimedia-on-demand server is reviewed. High data transfer-rate is emphasized. Design principle of disk arrays for video-on-demand or multimedia- on-demand server is presented. Size of striping units, data request size, data placement strategies and other important issues related with playback servers are discussed. Effectiveness of the idea is evaluated through the simulator raidSim.
05 Jun 1996