Re: accessing a kernel on a network
- To: mathgroup at smc.vnet.net
- Subject: [mg74124] Re: accessing a kernel on a network
- From: "Michael Weyrauch" <michael.weyrauch at gmx.de>
- Date: Mon, 12 Mar 2007 04:34:06 -0500 (EST)
- References: <esgpp0$l43$1@smc.vnet.net>
Hello, well, there is a "manual" method to attach a remote kernel to a local notebook , which does not require RSH or SSH. You just need to be able to somehow start up the remote kernel and enter commands in a remote notebook in order to get the connection going. This is described in some detail in http://support.wolfram.com/mathematica/kernel/configure/macintosh/unix22manual.html This webpage is somewhat old and written for Mathematica Version 2.2, but the information is essentially still valid for versions 5.x if adapted in obvious details. This I have checked recently, when I had a similar problem here. However, I consider this "manual" method only as a method for initial checks and tests of the connection, for everyday use it is somewhat awkward. It is much more convenient to set up the automatic startup of the remote kernel using the "kernel configuration options". In a sense, the problems, which can arrise here, are not Mathematica problems, but it boils down to be able to start up a process with all the correct parameters on the remote machine FROM your local machine. I suggest you experiment with this using a terminal on your local machine and send RSH or SSH commands to the remote machine. Then depending on the operating system of your remote machine you try to see what's going on there, i.e. if the mathematica kernel process starts up and if it was started with the required parameters. Note again: This first experimental step is completely outside your local Mathematica notebook! Only if this works convincingly then you should enter the successful command into the field "Remote shell command to startup kernel" in the kernel configuration options. Then you can configure a notebook to use this remote kernel, and whenever you open this notebook on your local machine the software fully automatically does everything that you can directly work using the remote kernel. If everything is done correctly you don't even realize that the calculations are actually done on the remote machine. (Note also: You can still also use the local kernel in another notebook in parallel, if you like to.) Good luck, Michael