On Tuesday 22 October 2002 2:56 am, Mark Weinstein wrote:
Antony,
Can you help me? I was reading your FAQ about the LinuxBIOS installation and had a few questions. Bear with me here... I am new to the whole LinuxBIOS thing and am having a hard time trying to find answers.
Hi Mark.
I see you asked Eric much the same questions, which have been discussed on the list overnight, however here are my answers to your questions, since I'm sure I too have a slightly different perspective than both Eric and Ron...
- I use Linux boxes in a clustering environment and would LOVE to have
the nodes boot themselves without having to load images from the network, etc. Is this doable?
You can certainly boot the kernel image from DoC for example, and you can then pick up your root file system from local HD, local Flash IDE, or the network.
However, what's your reason for not wanting to load the images across the network ? If it's speed or performance, I think you might be surprised how well something like etherboot can do...
- I use a RedHat 2.4.18 based kernel with quite a few kernel mods. How
much space is on the DOC (is it 8mb?) and will the OS fit on it?
Yes, the DoC I used is 8 megabytes - you typically need under 1 megabyte for the kernel image, so if you can fit the rest of your root fs into 7 megabytes you can build a totally self-contained (ie embedded) system.
I doubt this is the best way of setting up a cluster though ?
- If the answer to #2 is yes, then can you reflash the chips from a
server or some other machine and then have them reboot without taking the machines apart each time?
Yes, the fact that you programmed the DoC on the target motherboard in the first place means you could do it again once the system is running if you wanted to. Just bear in mind that if you get it wrong, you have a dead machine until you open it up and swap the Bios chip back in to reprogram your DoC with something known to work.
Changing the contents of the root fs on any sort of flash device is about as likely as reformatting and reinstalling your hard drive after booting from it and mounting it, though.
- If you can boot the OS from the chip, how do you have access to all of
your libraries, etc? Would you simply mount SOME file shares with libraries and apps via NFS?
You have three choices in effect:
1) Local flash "disk" storage - expensive and slow but it sounds cool
2) Local real disk storage - cheaper than flash, gets hotter and needs more power, but standard and plenty big enough
3) Remote mount across a network - no extra cost per node at all, shouldn't cause too much bandwidth problem depending on what applications you end up running, however don't just go for NFS because you've heard of it - it's really not that efficient - there are better network filing systems out there, but it's debatable whether there are any good ones :-)
Take a look at something like Tom's Root Boot to see how it's possible to create a workable system in a very small space.
I am VERY interested in building some servers with LinuxBIOS, but I am also very confused. :) Any help you can give me would be GREATLY appreciated!
Okay, now I'll tell you my interest in LinuxBios, whihc might help to put some of my comments in perspective.
I'm not (professionally) interested in clusters. I intend to play with clustering as a hobby someday, but for now I want to use LinuxBios:
a) for security, to prevent anyone from being able to boot a machine from floppy disk or CD Rom when I'm not looking
b) for speed, to get from power-on to login prompt (well, actually to NIC up and services running) as fast as possible
c) for novelty, so that when somebody switches one of my machines on it doesn't look like a standard PC about to boot Windows - it's obviously something different from the word go
d) (possibly) for cost - I have a particular application in mind which doesn't require much on the root fs, and I think it will fit into 7Mbytes on a DoC, so I may be able to build a useful "appliance"-type device without HDD at all
Hope this helps,
Antony.