Recently, both AMD and Intel added hardware virtualization to their CPUs. Now it's possible to write a Tiny Hypervisor that would trap whatever is necessary to distribute hardware resources to several OS, each running on its own VM. With some support from hypervisor, user-mode application under one OS can provide virtual device to another OS. What it means to development of a new OS is that it can rely on device support provided by another OS (Windows, Linux, or both): Device<-->OS#1<-->App<-->Hypervisor<-->OS#2.
In terms of general construction it's like microkernel OS, though logically it 's more like some kind of networking environment.
Currently, I'm working on hypervisor specifications.
Please e-mail me, if you're willing to participate in Hypervisor development _alexei@sbcglobal.net (yes, first character is underscore )
I would also appreciate your posts on my Tiny Hypervisor idea.

Thanks,
Alexei

Oh, another hypervisor coder in fasm board, welcome
Is your hypervisor going to be 100% fasm project?

such capable hypervisor would by no means be "tiny".

Yes
BTW, who's another one?


I expect it to be very small because of following reasons:
- It completely relies on CPU virtualization support.
- It will not include any hardware emulation code.
- No hardware (but CPU and memory) is shared.

Alexei

name "Hypervista" reveals the background
and welcome - I would gladly test this on my Cores!

This is one of the goals on my kernel, though its not entirely in FASM.

1)Object oriented kernel design supports multiple compile modes:
a) traditional end user GUI with applications and psudo c# compiler
b) server specific kernel with no overhead to perform server roles:
i) virtualization server kernel
ii) sql server kernel
iii) web server kernel
iv) any other server role can be compiled and built to run in stand alone mode with no overhead

I really dislike modern systems that require a gig+ of OS and applications to perform a single role (web server, sql server) and the machine sits in the closet and the gig+ of overhead is never utilized. Microsoft has some great server products (SQL server, IIS) but just by nature of having windows installed the stability and performance are compromised.

Greetings booter! I'm very happy to see another fasm board member interested in building a hypervisor. I'll be happy to help you out as much as I can.

I'm developing my fasm hypervisor as a Windows device driver specific to the Intel VMX instruction set. I'll get on AMD later. See this topic in the Windows section of the fasm board: http://board.flatassembler.net/topic.php?t=6504

vid (moderator) has stubbed out a fasm version of the VMCS (see the post near the bottom of this topic thread) http://www.board.flatassembler.net/topic.php?t=5748

I'm to the point where I'm building the VMCS and after that, I begin launching VMXON. Once that's done, I'll be including hypervisor based functionality and that's where the real fun begins. When I get to that point, I'm going to make an official Project announcment in the Projects and Ideas section of the board. I also want to seek to get some articles published and perhaps get on the agenda of some conferences to showcase the hypervisor and fasm.

You will find this board and the fasm community most helpful. In fact, Tomasz added VMX capability into fasm well before VMX capable processors hit the markets.

Best Regards.

Welcome to the fasm community.

Hi HyperVista!
Sorry for delayed reply (somehow, I missed reply notification).
My project (it's just an idea for now) takes different route:
- AMD64 X2 ("Pacifica" virtualization)
- "Bare metal" Hypervisor (boots before any OS)
- All providers of virtual hardware are user-mode applications on different VMs.
My main goals are:
- make development as simple as possible (no drivers )
- make the resulted system very flexible (no Host OS - any application can provide virtual hardware resource).
I hope to begin development later this year.
Thanks for your offer, now I know whom to ask for help!
Cheers,
Alexei

Actually the main security flaws are in IIS and SQL server, rather than the NT kernel

But sure, a more LEAN_AND_MEAN operating system would be nice when you don't need all the GUI and fluff for your server role. Perhaps one day tinykrnl and reactos will be suitable for that...