Over the past few months, Ivan Kelly has been working on porting MINIX to Xen as part of a university project. As last Monday was the final submission date for the project, he is now free to release it in the wild. Currently only the microkernel and the console driver have been ported, but a block device driver is in the works.
The code and a compilation guide can be found on the website. Contributors are welcome to join the project. On a related note, a new test version of MINIX has been released as well. OSNews reported extensively on MINIX 3 last year.
complication guide?
lol!
“a new open-source operating system designed to be highly reliable, flexible, and secure.”
So with all these “highly reliable” and “secure” operating systems around howcome Visa authorisation servers for online merchants are DOWN often?
Im tired of hearing about all these features that are not living up to their promises. Whatever happened to fault tolerence. We have lost hetrogeniou platforms in the networks and enterprise. This is because we cannot agree on the colour of shit. How long is it taking to get systems to talk together for petes sakes (Instant Messaging is a prime example here).
Doesnt this require code to be “modified” to support Xen? Isnt the purpose of a Virtual Machine that it is TRANSPARENT to the guest?
That’s just it, Xen isn’t a virtual machine like VMWare, it’s hypervisor. The idea is that you get more efficient use of the hardware and faster performance by tweaking the guest OSes a little. If you have a quad processor machine with 16gb of ram that you want to run 4 OS’s on, then feel free. For the rest of us, hypervisor allows better performance without having to get such beefy machines.
The performance difference between the Hypervisor and VM approaches IS significant.
So when we get hypervisor on hardware with the next wave of AMD and Intel chips and native OS support, where does this leave Xen? Just a Linux technology (and maybe OS X) only?
So when we get hypervisor on hardware with the next wave of AMD and Intel chips and native OS support, where does this leave Xen? Just a Linux technology (and maybe OS X) only?
No if you want to do virtualisation of the kind we’re talking about, several OS’s running at once without one as a “host” system, you’ll need a hypervisor. Basically what it does is sit there between your OS and the hardware controlling access to the hardware, especially cpu cycles, alotting them to the different OS.
Some nice graphics in this IBM research presentation :
http://www.cs.unm.edu/~fastos/05meeting/1
And a description of the Pseries hypervisor here: http://www.redbooks.ibm.com/Redbooks.nsf/RedbookAbstracts/tips0427….
Edit still doesn’t work. That first link was supposed to be : http://www.cs.unm.edu/~fastos/05meeting/2005-06-09-fastoshype.PPT
I can only assume that by native OS support you are referring to native Microsoft OS support since you mention OS X, and pretty much every alternative OS I can think of is using Xen.
Considering the fact that Xen has been implemented on actual systems and they seem to be quite a bit further ahead with a truly universal hypervisor than anything Microsoft has even hinted at (and MS does have a habit of advertising anything significant far ahead of actual releases), I would say that it is far more likely for someone to use Xen to run a Windows client with a Linux host in the near future than for anyone to be running a hypervisor created by Microsoft much less one that is provided through native OS support.
It is true that hardware virtualization support (provided by the next generation CPUs) will be necessary for Windows to be hosted successfully by Linux, but that is also because Microsoft hasn’t been willing to cooperate with an external hypervisor.
If it isnt a Virtual Machine , how does Xen give us the protection via isolation that VMWare would give for example?
Sorry for the spam on posts but I am unable to “Edit” my posts for some reason.
A VM is usually something like java or python/perl or QEMU. The point being that the program instructions are emulated/translated/mangled before being processed on the hardware. With Xen most of the instructions are run raw w/o any modification/translation. Memory access (perhaps other h/w stuff as well) is abstracted through an interface so that stuff isn’t really emulated/mangled/translated either. This means WAY superior performance.
Doesnt this require code to be “modified” to support Xen? Isnt the purpose of a Virtual Machine that it is TRANSPARENT to the guest or is it just the host that requires modification to support Xen? Maybe I missed something.
Here is the edited one, everytime I click “edit” i get “all of your text cannot be removed from your posts” blah blah. meh
Doesnt this require code to be “modified” to support Xen? Isnt the purpose of a Virtual Machine that it is TRANSPARENT to the guest or is it just the host that requires modification to support Xen? Maybe I missed something.
Yeah, but your hardware needs to support it. Otherwise the os needs to have the rights hooks to connect to the hypervisor which controls the access to the hardware.
For x86 based systems we’re slowly getting there with intel vt-technology and equivalents ( http://www.intel.com/technology/computing/vptech/ ) But it’s unclear if these extensions have been enabled in the latest intel chips like the core duo’s used in the iMacs.
http://en.wikipedia.org/wiki/Xen
So the hypervisor (Xen , Vanderpool and Pacifica) is basically the “host” and everything else is the “guest”. That is how I read it. Is this correct?
So the hypervisor (Xen , Vanderpool and Pacifica) is basically the “host” and everything else is the “guest”. That is how I read it. Is this correct?
Sort of. You still have a “priviledged domain” (domain 0) which can control the hypervisor, but it’s not a “host” in the sense that it virtualises the hardware.
To avoid confusing you (and myself) more I found a really nice explanation of how it works for Xen2 (no vanderpool) here: http://openlab-mu-internal.web.cern.ch/openlab%2Dmu%2Dinter… (see points 2.5 and 2.6)
From the page : http://wiki.xensource.com/xenwiki/XenArchitecture
Would attacks now target the hypervisor to be cross platform attacks now I wonder?
That seems possible, but unlikely. They would still have to go through the OS to get to the hypervisor in the first place which might cause a single attack to behave differently on different OSs, and the hypervisor seems like it would have few enough entry points that you could reasonably expect it to be pretty safe security-wise.
Someone correct me if I’m wrong, I really don’t know that much about Xen.
The more I think about it, the more it seems like an OS itself would have to be buggy (or compromised) in order to attack the hypervisor. So attacks would be extremely OS specific with no way to generically attack all of them.
Xen will use VT and Pacifica to run native OSes. Paravirtualised (Xen ported) guests will still run quicker than otherwise possible.
Even with VT / Pacifica, you still need a hypervisor to use the extra hardware capabilities.