Rad wrote on Jul 25^th, 2008 at 10:09pm:

Really, there isn't a clear boundary between any of these things, it's a spectrum. In this case they are running nested, but in a hypervisor (VMWare does most of its heavy lifting in kernel mode, not user mode).

If you believe in these classifications, VMWare is a type 2 hypervisor, whereas the old kind from the IBM 360 (VM/CMS) or the VAX VMM which were around 30-ish years ago is a type 1. The x86 architecture can't support the type 1 kind properly anyway; although originally designed for VM monitors in the multiple ring structure in the segmented days, that all fall apart when the 32-bit architecture and paging system was retrofitted on without enough thought.

What VMWare-type hypervisors do, though, is allow various kinds of hybrid operation; VMWare in a host OS does a mixture of hardware arbitration and emulation. What VMWare does to get speed from the emulation is do some nifty stuff - instead of typical privilege traps to emulation paths (which are excruciatingly slow on a lousy architecture like the x86), privilege traps cause the monitor to patch the faulting instructions with non-faulting code sequences so that the system running under VMware runs at pretty high speed without much emulation overhead.

And even in the old days the classic hypervisors, which started out mostly trying to arbitrate access to the underlying physical hardware, did end up doing outright emulation of some virtual devices. So, don't read too much into the categorization.

And these systems can run in different ways. VMWare ESX is pretty thin, so it's doing more hardware arbitration and less emulation, since there isn't really a hosting OS as such, making it in many respects it's closer to a classic type 1 hypervisor. VMWare under Windows or Linux or MacOS is doing a lot more hardware emulation, and that's more type 2.

Rad wrote on Jul 25^th, 2008 at 10:09pm:

Depends what you need, but probably very little. Since you'll probably be exploring web-type development, which isn't exactly resource-intensive, 512Mb would be more than plenty. Ubuntu has a nice performance monitor applet modeled on the Windows task manager, so you can learn what different activities cause in terms of memory load pretty easily.

Oh, and that Windows desktop you can see in the background is the host OS, which in this case is Windows (since I spend most of my time there, and Visual Studio really is a vastly nicer working environment than the others for C++ development).

Quote:

This errors when you try to launch PM 805 from install CD, which is bootable.

I first tried booting with the old hdd still in (testing my tools before needed), and figured it weird that it errored .. cuz maybe it didn't like seeing the installed copy.

But then when it errored upon booting up after I installed the NEW disk, uh, I couldn't figure that out .. cuz the new disk was empty .. except for the message I already posted.

Today I went back and added two more partitions, using PM from Windows.

1. Linux swap (1-GB)
2. FAT32 (1-GB)

both nested inside extended partition, at very end of drive, after shrinking the last (65-gig) partition 2-GB from its end.

One of the Ubuntu install guides I read suggested creating the FAT32 partition as a place where you can store downloads b4 scaning for viruses .. b4 importing these files into Windows/NTFS. I like that idea, and I did it in a way that I didn't have to create/use any more primary partitions.

I am thinking of busting the Linux proper ext3 into two .. so I can ALSO install a copy of CentOS 5, which the Rad server uses. Learning the server mojo is a big part of why I want Linux.

The big question tho > is 10 gigs too small for a Linux install?

I could do something like 12/8 .. if CentOS wouldn't need as much space.

That would mean using all four primary partitions. Anybody forsee a problem with that?

This question should probably be a new thread:

http://radified.com/cgi-bin/yabb2/YaBB.pl?num=1217103795/2#2