> Date: Wed, 24 Jul 2002 16:56:11 +0530 > From: "Adrian D'Costa" <[EMAIL PROTECTED]> > >> > I am being assigned a project for a school. They have a >> > tight budget. >> > Right now the are on a windows 2000 server and 28 nodes. >> > The server is P III 800 Mhz, 64 mb ram, 40 GB hdd and a >> > DSL connection >> > using a usb modem. >> >> Yeah, the server RAM will need to be *much* higher, >> especially if you are planning on users running big apps, >> like Netscape and StarOffice. For 28 clients, figure at >> the bare minimum 256MB and try to get 512MB. See how much >> the mobo can take. If RAM is cheap, just max it out. > > Thanks. I was thinking something like 256MB. I was also thinking of Clustering. Have > you tried it out?
I haven't tried clustering yet. I'm not sure it would relieve the memory crunch. If your server is swapping, however, your users will notice it in the form of slow performance immediately. 256MB will only be enough to get your first 10 or so clients online without swapping. To get the rest online without performance degradation you will need more RAM in the server. At least 512MB, but perhaps 768MB or 1GB. >> Also make sure the USB DSL modem is supported under Linux >> before you make the leap. Of course, *always* make sure >> your hardware is supported before switching to a new >> system... > > Well I have some doubts about the USB modem. Hardware is all Intel sytems. Mobo is > 830 I think. Red Hat posts their supported hardware lists on their website. If it's not supported explicitly, then just get another one and save yourself time and headaches. >> > The nodes P III, 450 Mhz, 64 mb ram, 40 GB HDD. Some >> > have cdroms some >> > scanners. All the NICs are Rl8193 >> >> Comfortable amount of RAM, but not enough for much local >> apps work. If you can double it to 128MB, that's a start. > > Again, is it possible to use the local harddisk or is there a way to mount them on the > server and utilse them? If you intend to use LTSP, then the terminals are not reliable storage units. The power could be cut at any time. You do not want to be using them for anything other than temporary storage and only for operations that influence the operations of the user sitting at that terminal. Yes, you could mount them with ENBD, but to what end? What data would you trust to be stored there? System data? User data? They're only good for swap under LTSP. As someone else suggested, you might *not* want to use LTSP if you feel you need to use these drives fully. Install Linux on the harddrive of each system and then figure out a way to manage them all centrally. No, I can't recommend this. This is the entire reason that LTSP exists. Maintaining separate complete OS installs on each machine is the management nightmare that has driven most of the people to LTSP. LTSP embodies the idea of centralized management. As little as possible goes on the workstations. You need to choose between ease of management or using all of your existing hardware. The two don't mix. >> > Their only request is that they need to utlise each >> > node's harddisk. >> >> only real use for hard disks in LTSP clients is for local >> swap, which you will definitely want if you want to do >> local apps. > > It this easy to setup up? Is it in the documentation? Easy? That depends on your skill level. If you're comfortable with bash scripting and command-line stuff, then it's not bad. Some people have posted things to the list. Not sure who or when. If you can't find the posts in the archive, you could try sending an explicit request to the list for such details and hope that someone feels like posting them again. I don't believe such things are in the documentation as it is not standard LTSP procedure yet. >> > They would be using Netscape, StarOffice, Java and the >> > usual stuff for >> > schools. I have some knowledge in Linux but not >> > successfully in >> > installing ltsp on my home systems. I think I can use >> > local >> > applications and would use each harddisk etherboot to the >> > server using >> > etherboot's lilo option. So I need you people's >> > assistance in what >> > would be the best way. >> >> I guess you could use the HDs for that and for swap. I >> don't understand what you're asking with regard to "the >> best way" though. Do you want details of how to set it all >> up? As far as I recall, local swap is not a simple option > > The "best way" to use the local harddisks as the client does not want it to go for > waste. Yes I need some way to set up local applications. Local applications in the LTSP sense means applications running in the RAM and CPU of the workstation. By definition, LTSP workstations are diskless and have no hard drive and therefore nothing is actually "installed" locally (they just run locally). "Local apps" is a standard part of LTSP and is well supported. This does not use the hard disks in the clients, however. I've heard of new mobos that allow up to 8 IDE connectors. You could build a couple of LTSP servers with IDE RAID and do load balancing between them if you wanted. Seach the list for "load balancing" for more on this topic. That would only use 16 of the 28 drives, though. You could take the remaining 12 and put them in other PCs which will not become part of the LTSP cluster and use them as auxiliary storage. >> in LTSP 3.0. You'll have to modify rc.local to probe for >> the IDE disk, setup the swap partition (or probe for an >> existing one), and then mount it and enable it. This has >> all been done before, but it's not standard yet. > > Any pointers where I can find it? Search the archives of this mailing list. Jason ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _____________________________________________________________________ Ltsp-discuss mailing list. To un-subscribe, or change prefs, goto: https://lists.sourceforge.net/lists/listinfo/ltsp-discuss For additional LTSP help, try #ltsp channel on irc.openprojects.net
