[nycbug-talk] Xen & FreeBSD

[H. G.]

My deepest apologies; my response to Pete's email didn't hit the list.
 Message below:

On 8/1/07, Peter Wright <pete at nomadlogic.org> wrote:
>
>
> > I was just wondering, has anyone managed to get FreeBSD running under
> Xen?
> > Anyone managed to get it to run as dom0, even?
> >
> > In my humble (and worthless) opinion, I think Xen would make an awesome
> > companion
> > to jails, since having the both of them means you have the option of
> both
> > "lightweight"
> > and "heavyweight" virtualization.
> >
> > So if anyone has managed to get this going, I'd be happy to hear about
> it.
> > :)
>
> I would not expect any para-virt bit's to get committed to the FreeBSD
> kernel any time soon.  There has been some work to get this going, but it
> was for the 5.3 branch.


:-( Thats saddening.

so - I wouldn't hold my breath on a domU implementation of FreeBSD.
> getting a dom0 is also probably not going to happen any time soon either -
> although the NetBSD team has had this working for some time.


I was using NetBSD to run Xen 2 guests for a while there.  Its pretty nice, and
the whole reason I would rather a BSD as a dom0 instead of Linux is because,
well, BSD feels much more "solid".  (Apologies for my n00bish
descriptive terms.)

I believe Net supports Xen 3 now too, but I was really interested in
running it on
FreeBSD.  But alas.... :-(

Here's my two bits (i've been doing alot of work with Xen and Jails while
> building HPC datacenters spread globally ) - i think both methods have
> their place.  Jailing works great in many environments where something
> like Xen would be overkill (core IT services come immediately to mind ).


Totally agree here.  Although wouldn't it be great just to run an OpenBSD
guest just for your infrastructure bits? :) (Although I do recognize that as

stupid since you're just increasing layers and attack surface for no good
reason.
"Just 'cause I can" is as good a reason as any, no? :] )

One of Xen's strengths is it's ability to set hard caps on memory and cpu
> usage, along with "live-migration"; although both potentially come with
> performance cost.  so really, i think they compliment each other.


Is there any work to bring these features to jails?  I've not been
around the BSD
universe for very long, but I'm 100% positive that _hordes_ of people would
be
absolutely thrilled at the idea of being able to jail CPU and memory
utilization.

I think before any of the *BSDs start tackling something as complicated as
> Xen i'd like to see better support for things like iSCSI (both hardware
> and software initiators/targets), FC and PXE.  when you get into
> virtualizing, the ability to decouple your storage from your CPU/RAM is a
> very important piece of this puzzle.


Agreed.

On a related note, haven't mainframes and big iron been working this
way already

for quite some time?  I have no experience with these, but the whole concept or
IBM's LPARS (if I'm not mistaken) is connected to this.  Its as if
tech on mainframes
and tech on plain-old COTS desktops and servers are converging upon each
other.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.nycbug.org:8443/pipermail/talk/attachments/20070801/4d436ef7/attachment.htm>