[tahoe-dev] detecting weak uploads and share rebalancing Re: Share rebalancing

[Brian Warner]

Zooko O'Whielacronx wrote:

> Ah yes, this is an example of why I wish for #302. If Tahoe-LAFS just
> used the standard Chord ring that all modern distributed data systems
> use, then you could figure out which shares go to which servers by
> visually inspecting the storage indexes and the server ids.

.. and we'd suffer from the "lumpy distribution" problems that are
discussed in ticket #302, where servers get unequal load depending upon
where they sit in the ring, and where servers who become full can dump
inordinate amounts of traffic on the poor node just clockwise from them.

And, an attacker who took out several neighboring-in-id-space servers
would kill or seriously damage several files (if they took out N-k+1
consecutive servers in a non-trivially utilized grid, they'd be
guaranteed to kill some files). In the permuted-ring design, they could
only easily target one file at a time, and taking out N-k+1 consecutive
servers would have no higher chance of completely killing a file than
taking out N-k+1 randomly-selected servers. There'd be no correlation
between share placement of independent files. (This property is closely
related to the non-lumpy-distribution issue).

I still believe that permuted-ring gives us better overall behavior. I'm
willing to be proved wrong, though :).

cheers,
 -Brian