What I would like to do is have a main computer support the operating system boot files and program files for running each node.
I have had people say you need to boot from network cards but thats all they could say, and provide no web links for the process.
In my opinion there should be possibly at least 3 ways.
One is the boot from nick card
Two is boot using a floppy and redirect you to the main machine.
Three I see stuff in the BIOS these days that I know nothing about for boot possibilities.
The reason I say machine or computer and not server is that I don't know for sure that server versions are required. I would 'like' to run it off of Win XP pro and the nodes Linux. The reason for MS XP pro is such things like explorer.exe for getting around file structures etc. (Way more Windows time.)
I had FREESCO hacked by somebody to run SETI but the loss of performance was horrable 20 to 30 percent longer times for SETI work units across Pentium 3 and Duron and AMD XP's platforms.
I have to go the house apraiser is here getting things lowered.
Back:
I have at the present time in my house enough stuff to make up to 14 processors run running in speeds from 700 mhz to 2,000 mhz.
Part of my problem is that I can only do it for part of the year cause I am in the southeast and the cost of airconditioning. The cat lives in the garage and the fur would clog and the heat would be a killer on hardware. (I can just see capicitors cooking off on older motherboards.)
Next is that comming across hard drives is an expense.
Your help is appreciated.
muttley
Last edited by muttley; 11-15-2002 at 03:51 PM.
DNAr was working on a headless *nix distribution system that required nothing but a floppy drive and a network connection for each node. His website:
http://dnaresearch.com.au/gahnix.html
Information on how to boot up diskless/headless systems via just a NIC is here:
http://etherboot.sourceforge.net/
Customizable ROM images to burn onto EPROM (for the NICs) can be found here:
http://rom-o-matic.net/
If you can get around in Linux, we got ya covered.
Heretic from Ars did an amazing write up about his research of netbooting for DC projects. He is the one who designed (and owns) the Prime Monster.
The set-up is well documented and it is (relatively) easily converted to other DC projects.
I know that Scott from MacNN has already taken Heretic's setup/configs/etc. and made them work with the Distributed Folding client, SETI and a few other projects and he had little working knowledge of Linux beforehand.
Unfortunately Scott didn't really document what he did, he just fiddled with it until it worked. You should be able to use Heretic's guides to get you very close and then adjust a few things from there to make it work for the DF client.
To boot from a NIC, you typically have to have a special boot rom, which you can burn yourself or which you can purchase (you buy a nic from bootrom places that has a preconfigured bootrom.
The image used to netboot (the image put in the boot rom on the NIC) is the same as what you use to boot from a floppy or vice versa. So you can use either option (or a mix of both, while testing it out).
Dnar's work looks very interesting, though I haven't tried it or talked to anyone who has. It would be the easiest route, if he has it working for DF.
http://ws9.jobnegotiator.com/html/homepage.html
That is the link for the Prime and SETI monsters.
Thanks, I'll start reading.
As for Linux I have only worked with Red Hat 7.2 and 7.3
I had DF running on it.
I couldn't reliably program Linux to have its own IP address cause it had like 7-8-9 lines to fill in and asked for Default Gateway, twice I think and I asked other people and they couldn't get it right for the fil in the blanks and I supplied the information I am use to seeing in windows.
like the cards IP address I wanted say 192.168.1.105
Subnetmask 255.255.255.0
Default Gateway 24.31.216.1
the ISPs Primary DNS and
the Secondary DNS
If I just chose use DHCP I had internet access fine but if I chose the the ip address I wanted the other information I would supply would be in the wrong order or something or maybe it wanted 128.0.0.0 some place and nobody had the right information.
I tried making corrections inside Red Hat but it didn't work and I tried dozens of times reloading Red Hat and that is just such a slow process.
muttley
I think that the primary problem Scott had was getting the server stuff figured out.
I will try and talk him into dropping by here to offer any advice he can.
Thanks,
(once again Red Hat is just an example Linux.)
Even if it takes a lot of memory to say put up a Linux/Red Hat machine I still wish to do it.
I can whittle down the active memory space some cause I don't plan on needing all the standard loaded applications Linux/Red Hat installs.
On each node but I don't see why I can't over a 10/100 nick card be able to use the hard drive in the server having a /swap file if I get into trouble. I'm willing to swap hard disk speed access and still get it up and running.
muttley
Naw, you can run the whole thing out of RAMdisk, the hard part is getting it to netboot.
Well, to avoid a lot of data loss if something crashes, you want to write to the results to the server every so often. I am no expert though, I just read what they say when they are talking bout it.
Most of it is over my head.
I will try to holler at Scott again next time I see him on IRC.
Thanks MAD-ness and Halon50
muttley
It might not help you directly, but you might be interested to hear that a few of us over at BroadBandReports are working on Project BOUNCE:
http://project-bounce.sourceforge.net/
I can see you wondering...BOUNCE? right..what does it do
The goal is to:
"Create an Open Unified Network Computing Environment intended for existing and future Distributed Computing (DC) clients, but with unlimited possible uses. The open framework will allow for many current and future DC projects. The BOUNCE client and interface will allow for remote browser based management of diskless & headless client arrays sometimes known as "farms". The key point to the project is to make the setup and maintenance of such an array such a simple a trivial task that a newcomer to Linux wouldn't hesitate to start using Linux."
Licence: GPL
We don't have a first release yet, but we're working hard to get that one out as soon as possible.
(And yes, the Distributed Folding client is already running)
If we succeed, managing a farm of diskless nodes is going to be easier then ever before
And yes, there are some screenshots available on our website...hopefully those will give you an impression
*sigh*
Wish I had the equipment to build a cluster of my own... would bang together a custom (686 optimized) Linux distro and a system for starting/stopping/checking progress/ of the DF clients, accessable by http... that'd be sweet!
Tip, check out www.linuxfromscratch.org! If you are novice at Linux, that's a great start to become somewhat more indepth.
pmfp sends
"There's only one bullet with your name on it, there's a thousand other ones that are addressed 'To Whom It May Concern'."
-Tracy Paul Warrington
Well... sort of. LFS isn't a bad distro to learn stuff from, but you don't learn much about Linux just by following the stuff in the LFS book. What you really learn from is reading through the packages you install, and figuring out how to change things. Fixing stuff that goes wrong also helps quite a bit.
It helps more than blindly typing in the stuff in the LFS book, at least...
Btw, pretty nice interface... although I am perplexed of the reasons for what seems to be the chosen solutions for some tasks (it's probably just me, who interprete the screenshots incorrectly).Originally posted by Starfish
It might not help you directly, but you might be interested to hear that a few of us over at BroadBandReports are working on Project BOUNCE:
http://project-bounce.sourceforge.net/
I can see you wondering...BOUNCE? right..what does it do
The goal is to:
"Create an Open Unified Network Computing Environment intended for existing and future Distributed Computing (DC) clients, but with unlimited possible uses. The open framework will allow for many current and future DC projects. The BOUNCE client and interface will allow for remote browser based management of diskless & headless client arrays sometimes known as "farms". The key point to the project is to make the setup and maintenance of such an array such a simple a trivial task that a newcomer to Linux wouldn't hesitate to start using Linux."
Licence: GPL
We don't have a first release yet, but we're working hard to get that one out as soon as possible.
(And yes, the Distributed Folding client is already running)
If we succeed, managing a farm of diskless nodes is going to be easier then ever before
And yes, there are some screenshots available on our website...hopefully those will give you an impression
Good luck with your efforts.
"There's only one bullet with your name on it, there's a thousand other ones that are addressed 'To Whom It May Concern'."
-Tracy Paul Warrington
TrueOriginally posted by bwkaz
Well... sort of. LFS isn't a bad distro to learn stuff from, but you don't learn much about Linux just by following the stuff in the LFS book. What you really learn from is reading through the packages you install, and figuring out how to change things. Fixing stuff that goes wrong also helps quite a bit.
It helps more than blindly typing in the stuff in the LFS book, at least...
"There's only one bullet with your name on it, there's a thousand other ones that are addressed 'To Whom It May Concern'."
-Tracy Paul Warrington
More to fiddle with is always good. I've actually taken my LFS 3.3 packages, married them to the LFS 3.0 bootscripts (I like the 3 digits of order information in the symlinks with SysV init -- newer than 3.0 used only 2 digits, so the symlinks looked like S70apache, but I like S700apache), and changed a bunch of other things as well (like the setup of shadow, etc.). I've also obviously added a bunch of stuff to the distro after installation -- X, Mozilla CVS, and some games (oh yeah, and mplayer recently, for the AVI Dragonball (
Thanks for the information I'll look into bounce.
In the mean time I have ran across LTSP.org and K12LTSP.org
From what I can tell this is what I am looking for.
For me it can work on 2 levels.
One, is the ability to demo that a machine can run on 32 megs of memory and off load the processing to a higher powered server and the server can be a work station. (the station needs no hard drive.)
Two, is the station given more memory can operate using its own processor and memory and be able to do DC also. Thus you can overclock a ie duron 700 to say 927 and all you have to get functioning is the Video and the LAN.
On a side note a person can move from one station and sign in and all the users preferences are just transfered to the next station. Hence only the hard drive on the server to boot and provide the necessary components.
muttley
Starfish: you guys should make sure to check out the work of Heretic of Prime Monster fame and dnar (GAHnix, etc.) as they both have done a lot of work that could come in handy.
I have played with Wayne's (aka dnar) GAHnix and it works really well. The instructions are pretty easy to follow as long as you have some basic *nix knowledge.
http://dnaresearch.com.au/gahnix.html
Train hard, fight easy
We are indeed inspired by the work of Heretic and dnar...fascinating stuff !Originally posted by MAD-ness
Starfish: you guys should make sure to check out the work of Heretic of Prime Monster fame and dnar (GAHnix, etc.) as they both have done a lot of work that could come in handy.
Thank you for the interface complimentOriginally posted by pmfp
Btw, pretty nice interface... although I am perplexed of the reasons for what seems to be the chosen solutions for some tasks (it's probably just me, who interprete the screenshots incorrectly).
Good luck with your efforts.
I'm not sure however why you are perplexed about the chosen solutions for some tasks.
If you're willing, I'd appreciate it very much if you could extrapolate a bit on this. (my e-mail addy is half way down the BOUNCE homepage)
I must say that I'm still quite a beginner regarding *nix platforms, netbooting etc..
But we do know what we want: a good overview/management tool for netbooting farms.
Perhaps we might not get everything directly right, but we're doing our best to create something functional, good-looking and cool for the entire DC community....so feedback is always very welcome!
We will certainly try to come up with a better description and information about BOUNCE as soon as possible.. but I'm also in the final year of my study and quite busy with many, many other things, so progress might not always come in a steady and predictable way
Posting Permissions
Reply With Quote