First , thanks , Marcel
If I got Mark right, it's not the image type qcow itself, he's
but the COW (Copy-on-write) feature.
You did indeed got me right; I am not talking specifically about
some format like qcow itself but my question is general. As I said
before, it seems that there are many
Copy-on-Write solutions; at least according to the Xen wiki page on COW
For example, Xen blktap, DmUserspace ,UnionFS, BenRin's Copy-on-Write
md device,Translucency, parallax, CowNFS , and LVM snapshots (and yet
I don't know yet whether MarkW XenFS should support cow).
I am not an expert on CoE. As I understand, there are two main
categories for Copy On Write solutions: file based solutions and block
It's obvious to me that there are some pros and cons for each method.
It seems to me that no doubt that when using a solution which should
patch the kernel
this can be thought of as a drawback by many. Also when the solution
is tightly coupled to a filesystem (like unionFS or CoWNFS) this maybe
can seem a drawback. (conmparing to block device solutions). So if
somebody can give a short analyze of pros and cons of the different
CoW techniques available for Xen
and the roadmap ahead this can make things better understood.
On 3/6/07, Ian Pratt <m+Ian.Pratt@xxxxxxxxxxxx> wrote:
> > I've been using qcow images generated by qemu-img for Linux and
> > MSWIndows HVM guests, and I've been getting good results. The image
> > files are much smaller that the equivalently sized raw images.
> > I simply create them (via gemi-img create -f qcom 10G) and use the
> > disk=['file:...'] line in my guest configuration files.
> If I got Mark right, it's not the image type qcow itself, he's
> interested in,
> but the COW (Copy-on-write) feature. Some time ago I tried to get
> COW working with qcow images, without success. So I guess the name
> is missleading, and (in Xen) there's no COW based on qcow.
> I'd really like to get a simple file based COW - is this planned for
> future Xen releases? Maybe someone can comment?
AFAIK it works in xen-unstable, so will be in 3.0.5.
Xen-devel mailing list