[Box Backup-dev] Re: Learning from ZFS (fwd)
Ben Summers
boxbackup-dev at fluffy.co.uk
Thu May 10 12:46:23 BST 2007
On Thu, 10 May 2007 11:28, Martin Ebourne wrote:
> Wout Mertens <wmertens at cisco.com> wrote:
>
>> On 09 May 2007, at 18:25, Martin Ebourne wrote:
>>
>>
>>> On Wed, 2007-05-09 at 16:58 +0200, Wout Mertens wrote:
>>>
>>>
>>>> 4. One thing that would really rock if extra code were added that
>>>> only stores blocks with the same checksum once. ZFS currently
>>>> doesn't
>>>> have that, but I think it's technically feasible to do something
>>>> like
>>>> that.
>>>>
>>>
>>> Er no, that wouldn't rock at all!
>>>
>>
>> Care to elaborate?
>>
>
> Well obviously given two arbitrary blocks that have the same checksum
> (or sha hash or whatever), it's very unlikely that the blocks are
> actually the same. It's pretty important for a backup system to give
> you back the actual data you stored, not just some data that happens
> to have the same checksum!
Hmmm. Yes and no.
Yes, you want your original data back. 100% guaranteed.
No, in that if you stick to this rule absolutely you can't use rsync
or Box Backup's rsync-like algorithm.
Maybe, in that there's a lower chance of it being a problem in the
rsync case.
Maybe we should add an option to turn off bandwidth efficiency?
Ben
More information about the Boxbackup-dev
mailing list