Re: [Vserver] VServer vs OpenVZ

From: Herbert Poetzl <>
Date: Wed 07 Dec 2005 - 16:57:09 GMT
Message-ID: <>

On Wed, Dec 07, 2005 at 05:57:31PM +0200, Alex Lyashkov wrote:
> ? ???, 07.12.2005, ? 16:34, Herbert Poetzl ?????:
> > On Wed, Dec 07, 2005 at 06:44:43AM +0200, Alex Lyashkov wrote:
> > >
> > > > (will use Z for OpenVZ and S for Linux-VServer)
> > > >
> > > > > Factors of interest are
> > > > > - stability,
> > > >
> > > > Z: the announcement reads "first stable OVZ version"
> > > > S: we are at version 2.0.1 (> two years stable releases)
> > > >
> > > And all this time VServer need a hack for allow bind socket to
> > > INADDR_ANY at VPS ;-)
> >
> > hmm, well, it works reasonably fine .. no?
> >
> The thing that I find unreasonable is that you cannot bind to
> INADDR_ANY on the host server, without affecting all the VPSes. This
> basically means that if you have default bind (or ssh) installed on
> the HOST system, no VPSes will be able to bring up those services.

no, you are wrong there, the difference is you _can_
bind to INADDR_ANY on the host and it _will_ do the
expected thing (i.e. bind to _all_ configured IPs)

of course, you also have the choice to put the entire
host into some chbind jail, so that binds to INADDR_ANY
are restricted to a subset ...

so, IMHO it's a feature ...

> > > Z Have: COW VFS (simular unionfs),
> >
> > S: has CoW link breaking (more powerful than unionfs :)
> >
> CoW link requires glibc modification, while VZFS does not.

if it does, then you're doing something wrong ...

(JFYI: the CoW link breaking is done in the kernel)

> The negative side of VZFS is double buffering, due to it
> being stackable FS.

yes, that's why we avoid it ...

> > > Virtual networking support,
> > > Fair-share scheduler
> >
> > S: priority and hard cpu schedulers
> >
> This is not the same. Priority and hard limits don't provide
> fair share CPU scheduling. Two different things (three).

well, 'fair share' has been in linux-vserver for ages

> > > Z don`t have: disk namespace support
> > > (but create vroot simular FreeBSD Jail).
> >
> > but I guess we are heading towards a feature
> > shootout (which is fine for me, but isn't very
> > important for the userbase, I guess :)
> >
> > best,
> > Herbert
> >
> > PS: nice to hear from you!
> > PPS: is there anything left from the cooperation
> > we started a year ago (or so)?
> >
> It would be great, but somewhat difficult as Vserver and FreeVPS use
> different ideology. VServer and OpenVZ store context id at kernel
> objects level (it's easier to develop, but slows down access to
> context data).

> FreeVPS stores pointer to context structure, at kernel object (more
> work, debugging, but much faster access to data, that is needed for
> each context switch).

hmm, again JFYI, linux-vserver uses both, context references
as well as xid information, wherever the one or the other
applies ...

btw, do you have any test results showing that FreeVPS
contexts are faster? if so, please provide them ...

> I see a lot of potential in the possible merge, but I consider the
> way FreeVPS works with context ids to be much more efficient. I also
> consider it fairly critical for optimum performance on a server with
> large number of VPSes. It would be great to work together though, but
> it makes sense only if VServer developers are interested in merging
> projects/porting some of the FreeVPS staff to VServers as well.

cooperation is not about porting stuff from one project
to the other, cooperation is about discussing stuff and
ideas, and keeping the APIs (and maybe userspace tools)
somewhat compatible, so that folks (the customers) can
choose to their likings ...

> Other side - FreeVPS used RH EL kernels. this adds stability to the
> kernel API, while all the bug fixes are back ported. At the same time

I do not understand why folks always think that back
porting bug fixes adds to stability ... well, maybe
to the stability of the API but not to the kernel
stability ...

> Vserver can continue to use bleeding edge kernels/be portable across
> all the platforms If compare features - FreeVPS has all features which
> VServer have at x86 platform. Yet, at this moment it is the only
> platform supported.

hmm, looking here I see a bunch of differences:

btw, is there a patch (or even better, broken out
version) available which applies to mainline?
(just curious)


PS: please point me to _any_ patch against a well
known base kernel (for the recent FreeVPS versions)

> --
> FreeVPS Developers Team
> Positive Software
Vserver mailing list
Received on Wed Dec 7 16:57:33 2005

[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Wed 07 Dec 2005 - 16:57:37 GMT by hypermail 2.1.8