Abstract: . . .  ‘kickstart’ helps a lot against such a situation. However with it you have to install a system from scratch at each node PC. (you must specify which application packages (RPMS) you need, edit configuration files and so on) In some PC cluster sites, people make dead copies of hard disk partition image of a node in which a . . .
. . .  duplex switching hub , it achieve maximum performance. For example, 2GB disk partition copy among 8 nodes over 100BaseT took less than 5 minuntes. By the pipelining and multi threading, dolly has very good scalability in increasing the number of nodes. Figure.1 shows expected elapsed time for copying against the number of nodes.  . . .
. . .  kickstart, automatic and very fast system installation 1 and upgrading could be performed. Keywords: Linux Cluster farm installation 1 Background Today, the number of nodes in a PC cluster is increasing and increasing, even in the situation where newly developed single CPUs are becoming more . . .
. . .  4MB chunk size,7MB/s transfer speed 4GB cloning expected curve 2GB cloning expected curve measurements Figure 1: time needed for disk cloning iofiles 3 <- # of transfer file /dev/hda1 > /tmp/dev/hda1 <- hard disk image /data/file.gz >> /data/file . . .
--1371,4,171,1648,6854