> From:
>
> Joseph Mocker <[email protected]>
> [email protected] wrote:
> > Quoting Joseph Mocker <[email protected]>:
> >
> >> A Hettinger wrote:
> >>> one method that does work, is to use a comstar iscsi port on the
> >>> loopback device, mount it locally then use "dd" to move the content
> >>> of your old backing store to the new device.
> >>>
> >>> you can then set up views to whatever you want. it's ugly, but it
> >>> does work.
> >>>
> >> Ah, clever. hopefully a sparse ZVOL would be smart enough not to
> >> allocate blocks of zeros.
> >>
> >
> > I'd hold off on the sparse volume. I've done some testing that seems
> > to suggest that sparse volumes *seem* to be less performant than
> > regular zvols.
> >
> > I'm compiling the data now and I hope to publish it to the list
shortly..
>
> Really? But non-sparse volumes just reserve space, but my understanding
> is that otherwise they don't really work differently internally. They
> both allocate off a free list. I think the COW methodology won't ensure
> that ZVOL blocks are actually on contiguous blocks of media. That's very
> surprising to hear.
>
> --joe
In my tests I saw performance issues with using zvols at all. I ended up
giving up and just using files.
(admittedly, I think I gave up around 107)
.02 USD
A Hettinger
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
