From owner-freebsd-questions Mon Mar 2 08:37:01 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id IAA05449 for freebsd-questions-outgoing; Mon, 2 Mar 1998 08:37:01 -0800 (PST) (envelope-from owner-freebsd-questions@FreeBSD.ORG) Received: from mhv.net (root@spice.mhv.net [199.0.0.21]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id IAA05334 for ; Mon, 2 Mar 1998 08:36:40 -0800 (PST) (envelope-from mgraffam@mhv.net) From: mgraffam@mhv.net Received: from localhost (port77.mhv.net [206.229.41.5]) by mhv.net (8.8.5/8.7.3) with SMTP id LAA02559; Mon, 2 Mar 1998 11:36:25 -0500 Date: Mon, 2 Mar 1998 11:29:42 -0500 (EST) X-Sender: phundie@localhost To: Kevin Brown cc: freebsd-questions@FreeBSD.ORG Subject: Re: FreeBSD in a multi-processor/machine array In-Reply-To: <3.0.32.19980302091214.00acc320@mail.primelink.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Mon, 2 Mar 1998, Kevin Brown wrote: > I am beginning to do some procurement of smaller 286, 386, and 486 machines Hold this thought. > 1) Maintain one central server (dual pentium) with full FreeBSD source. This shouldn't be a problem. Just set up the machine and archive the source, nfs export the directory to the smaller machines and you're golden. > 2) Allow the smaller machines to boot their operating system from the > central server utilizing perhaps only a floppy drive. This should not be much of a problem either. I have not personally used FreeBSD in this fashion, but I have set up Linux this way, and I'd imagine what Linux can do in this area, FreeBSD can do as well. You can easily mount your /usr via NFS. Root NFS is a little weird, but it isn't too difficult. At the very least, a floppy or ram disk could be your root and /etc and stuff could be mounted from an nfs drive somewhere. > 3) Provision the machines so that they rely upon the mass processing > abilities of their neighbors to complete tasks. Yes and no. The short and quick answer is no. You can't run the machines in parallel to complete tasks like a shell script or even to play doom or quake. The long answer is that you can use PVM (Parallel Virtual Machine) to create a virtual machine for parallel execution of tasks over the network. These tasks, however, would need to be compiled especially for PVM and would need to use PVM's libs for stuff. > 4) Where applicable share drive space across an Ethernet or Fast Ethernet > connection. This goes hand in hand with NFS mounting of root above. > 5) Allow remote users to login via telnet or ssh to the central server and > have the ability to harness the full processing ability of the array. > Without having to play magical chairs on logging into multiple machines. This goes hand in hand with the question of PVM. Again, the short answer is no. The long answer, if you are going to write your own code is absolutably. On the other hand.. if you are a fairly good programmer you could set up a system like this: have all drives shared via NFS. Use rsh or ssh (I'd use the later) to run commands on remote machines, ie ssh host1 mathematica & would fork mathematica to your X display. Now, in your situation with 40 or so machines, you wouldn't want to have to remember which machines are under what load. It should be possible to develop an interface to the network (it could probably even be a hack to a standard Unix shell) that would find out what physical drive the program you want to launch is located on and contact that host. It could then check the load average for the machine, if it is under some N percentage, that machine would launch it, else the interface tries another machine. Properly implemented, this would give you the ability to just type "mathematica" into your modified bash (or whatever) and then the system could/would launch it on the machine that would give the best performance. The code wouldn't exactly be trivial however. It would require a good bit of network oriented code, and probably a tiny bit of systems programming as well. Good stuff though. If I ever had 4 or 5 machines that I could play with like that, I'd try to get it working. Michael J. Graffam (mgraffam@mhv.net) http://www.mhv.net/~mgraffam -- Philosophy, Religion, Computers, Crypto, etc "..subordination of one sex to the other is wrong in itself, and now one of the chief hindrances to human improvement.." John Stuart Mill "The Subjection of Women" To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message