----------------------------------------------------------------------------- Distributed Maple application software INSTALL file (c) 1998-2000, RISC-Linz, see file COPYRIGHT http://www.risc.uni-linz.ac.at/software/distmaple $Id: INSTALL,v 1.3 2000/06/16 13:26:05 schreine Exp $ ----------------------------------------------------------------------------- ----------------------------------------------------------------------------- Note for RISC-Linz users ------------------------ Ignore steps 1-5, everything is already set up for the machines of our network. Start with step 6 where DISTMAPLE is "/home/info/www/software/distmaple" DISTSOFT is "/home/info/www/software/distmaple/distsoft/distsoft" In Step 6c, you need NOT modify the file "dist.systems" copied from DISTMAPLE. The key "octane" can be used for our SGI Octanes ("andromeda", etc), the key "linux" can be used for our Linux PCs ("thetys", etc). ----------------------------------------------------------------------------- 0) General ---------- The installation constists of the following steps 1) Install Maple 2) (Optional) Install CASA 3) Install Java 4) Install Distributed Maple 5) Install the 'distsoft' directory 6) Setup your working directory 1) Install Maple ---------------- Install Maple V.5 or higher on every machine to be used, see http://www.maplesoft.com The command maple must work (or an equivalent command by which Maple can be invoked). 2) (Optional) Install CASA -------------------------- If you want to use the parallelized CASA functions, you have to download and install CASA on every machine to be used, see http://www.risc.uni-linz.ac.at/software/casa If you start Maple, the command with(casa): must work. If you don't have CASA, you can still use the parallelized versions of the standard Maple functions of this software. 3) Install Java --------------- Install Java 1.1 or higher on every machine to be used, see http://www.javasoft.com The command java must work (or an equivalent command by which Java can be invoked). Note: If you use Linx PCs, only use a Java version that supports native threads (such as Java 2). We experienced severe problems (long blocking tasks) with green threads under Java on Linux. 4) Install Distributed Maple ---------------------------- Download the Distributed Maple package from http://www.risc.uni-linz.ac.at/software/distmaple uncompress and untar it in a global network directory that can be seen by every machine (or, alternatively, install a duplicate on every machine that has not access to the global network directory). Let DISTMAPLE be the name of this directory, i.e., if you have created a directory /home/info/www/software/distmaple dist.systems dist.maple dist/*.class ... then DISTMAPLE is "/home/info/www/software/distmaple". 5) Install the 'distsoft' directory ----------------------------------- Download the 'distsoft' software from http://www.risc.uni-linz.ac.at/software/distmaple uncompress and untar it in some directory that you can see on the machine from which the parallel computation is to be started. Let DISTSOFT be the name of this directory, i.e., if you have created a directory /home/info/www/software/distmaple/distsoft README (this file) init.maple dist casa maple ... then DISTSOFT is "/home/info/www/software/distmaple/distsoft". 6) Setup your working directory ------------------------------- Go to the directory from where the parallel computation is to be started. a) Create a symbolic link to the Distributed Maple installation: ln -s DISTMAPLE/dist dist ln -s DISTMAPLE/dist.maple dist.maple You must see the files "dist/*.class" and "dist.maple". Check whether ls dist/*.class more dist.maple work. b) Create a symbolic link to the 'distsoft' installation: ln -s DISTSOFT distsoft You must see the file "distsoft/init.maple". Check whether more distsoft/init.maple works. c) Copy "dist.systems" from DISTMAPLE to your working directory cp DISTMAPLE/dist.systems . and adapt it to your network. For each different type of machine create an entry like KEY rsh java -classpath :DISTMAPLE dist.Scheduler maple -q `dist/isclient` := true: read `DISTMAPLE/dist.maple`: 14 0 0 as described in the manual of the Distributed Maple installation: - KEY is an arbitrary string by which the kind of machine is identified. - "rsh" is a command by which you can start a process on that machine, check whether rsh ls works. - "java -classpath :DISTMAPLE dist.Scheduler" is a command by which the Java program "dist.Scheduler" is started on the machine. You have to add in the "-classpath" option the DISTMAPLE directory and all other directories on which the Java installation depends. Check whether on the denoted machine java -classpath :DISTMAPLE dist.Scheduler works and gives an output starting with the line options: [-b] [-n] arguments [ ] ... If you have different Java or Distributed Maple installations on different machines, you have to make for each an entry in "dist.systems" with the corresponding settings for and DISTMAPLE. - "maple -q" is the command that invokes Maple in quiet mode. Check whether on the denoted machine maple -q works. This command does not give any output, type quit; to terminate the program. If you have different Maple installations on different machines, you have to make for each an entry in "dist.systems" with the corresponding setting for DISTMAPLE. - "`dist/isclient` := true: read `DISTMAPLE/dist.maple`:" Insert the appropriate value for DISTMAPLE. If you have different Maple installations on different machines, you have to make for each an entry in "dist.systems" with the corresponding setting for DISTMAPLE. - "14": a speed parameter, higher values means faster machine. Create one entry for each kind of machine, machines with higher numbers will be preferred for execution. - "0 0": use this (see the Distributed Maple manual for further info). ----------------------------------------------------------------------------- $Id: INSTALL,v 1.3 2000/06/16 13:26:05 schreine Exp $ -----------------------------------------------------------------------------