The file(s) will be read from the fileserver for each copy to a node, resulting
in N hits on the fileserver, where N is the number of nodes in the machines file,
as seen in the black dashed arrows in the diagram below. This method works, but
soon runs into scalability issues. An alternative method is to use the
mcp utility, which will copy the file from the fileserver once, then perform
the remainder of the copies in a tree fashion from compute node to compute node,
as seen in the green solid arrows. To use mcp, the final line in the setup script
should read:
REM Copy the executable (and data files, if necessary) to each node
REM The next three lines should be entered as a single line command.
\\tc.cornell.edu\tc\CTCToo~1\mcp.exe
\\tc.cornell.edu\tc\users\your_userid\lab\wavesend.exe
t:\your_userid
