J.M. Rosario and A.N. Choudhary, "High-performance i/o for massively parallel computers: Problems and prospects," Computer , pp.59-68, Mar. 1994. Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
An Efficient Delay-Optimal Distributed Termination Detection.. - Mahapatra, Dutt (Correct)
....factors that make detection delay the most important metric to optimize. First, detection delay signifies a waste of computing resources of the parallel or distributed system. This is accentuated by the fact that the gap between processing speed and communication speed is widening every year [18, 20], and since detection delay is primarily dependent upon inter PE communication delays, for the same detection delay, more computing cycles are wasted in newer systems. The situation is especially severe in a distributed system of, say, workstation clusters connected by a local area network (LAN) ....
J.M. Rosario and A.N. Choudhary, "High-performance i/o for massively parallel computers: Problems and prospects," Computer , pp.59-68, Mar. 1994.
Job Scheduling in Multiprogrammed Parallel Systems - Feitelson (1997) (16 citations) (Correct)
....are computation intensive. Some require significant amounts of I O. In the past, such applications were rare, because parallel machines did not have the I O facilities to support them. But as parallel I O mechanisms are developed and gain recognition, more and more I O intensive jobs will appear [153, 202]. The system will then benefit from the possibility of overlapping the I O of one job with computation of 79 another, just as in uniprocessors. This can be based on collective I O operations 23 and gang scheduling. As in uniprocessors, jobs that perform I O should be given higher priority when ....
J. M. del Rosario and A. N. Choudhary, "High-performance I/O for massively parallel computers: problems and prospects". Computer 27(3), pp. 59--68, Mar 1994.
Clustering and Classification of Large Document Bases in a.. - Ruocco, Frieder (1997) (3 citations) (Correct)
....of the document vectors is more complex. The primary 7 concern was that each processor must access the same files at the same time. This lead to significant problems with file I O. It is known that file I O can be a system bottleneck and that performance for I O is highly system dependent [6]. There are several other concerns with file I O, such as system load, over which the user has very little control. The Intel Paragon offers several methods for file access. These are categorized based on the level of synchronization inherent in each mode. A detailed description of the performance ....
J.M. del Rosario, A.N. Choudhary, "High-Performance I/O for Massively Parallel Computers," IEEE Computer, March 1994, pp 59-68.
The Scotch Parallel Storage System - Gibson, al. (1995) (16 citations) (Correct)
....in these environments remains an unanticipated occurrence, their storage is managed as a unit and may be endowed with RAID functionality. In many environments, these fast client machines are used for time consuming computations such as VLSI simulation, weather simulation, and rational drug design [13], whose datasets are often massive (10 MB 100 GB) With the wide availability of high level parallel programming tools, such as PVM, high performance FORTRAN, and distributed shared memory (DSM) there is a growing trend to implement each of these appli Figure 8: Visualizing 3 D dataset slices ....
del Rosario, J.M. Choudhary, A.N. "High-performance I/O for Massively Parallel Computers: Problems and Prospects, " Computer, 27(3):59-68, 1994.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC