[tahoe-dev] measure your convergence
zandr at allmydata.com
Fri Mar 21 05:39:18 UTC 2008
On Mar 20, 2008, at 1:38 PM, zooko wrote:
> Ever wondered how much storage space you would save if you and your
> friends coalesced all of your identical files?
In thinking about the implications of the loss of universal
convergence, a few things come to mind. I was chatting with Zooko over
IM about these, and he suggested I share them more broadly.
Though I haven't run dupfilefind on my own machines yet, intuition
tells me that I don't really have a lot of duplication on a single
system. However, convergence is still important, because while I'm not
likely to have lots of local duplication, it *is* likely that
something will have happened to cause me to forget that I've already
uploaded something. Convergence accelerates these otherwise wasted
Further, broadening the scope of the chk_secret to all of a user's
machines will indeed offer some savings. .Mac is actively ensuring
that there are a few GB of files that are duplicated between my
notebook and desktop machines, and as they're both Macs, there are
likely very large overlaps between the two boxes.
Also, having a single chk_secret per account paves over a number of
opportunities for errors in restore-then-backup cases, where a machine
recovered from the mesh would generate a new chk_secret and then re-
upload its entire contents.
More information about the tahoe-dev