Re: [bailey-developers] lattice master

[Ning L. <nin...@gm...>]

On Fri, Feb 29, 2008 at 6:54 PM, Doug Cutting <cu...@ap...> wrote:
> Ning Li wrote:
>  >>  You've switched to counter-clockwise replication, but I think that's
>  >
>  > Oops, I always used clockwise replication. :)
>
>  Just to be clear: when a new document D arrives, the question is what
>  nodes on the ring do we send it to?  The N clockwise (greater) or
>  counterclockwise from (less than) D?  The most recent messages assume
>  counter-clockwise (less than).  A document between C and D would be in
>  the indexes on nodes A B and C.  The Dynamo paper says "clockwise", but
>  I think that direction makes things more confusing and that we should
>  instead use counterclockwise.  Note that the index at node C contains
>  documents in range C-F, three nodes clockwise, but the Dynamo paper uses
>  the term clockwise when talking about assigning instances to nodes, not
>  the range that each node contains.  Are we in agreement about this?

I'm confused with which is clockwise and which is counter-clockwise now. :(

I thought node C serves D-E and E-F in addition to C-D, so it's clockwise.
You are saying because range C-D is on nodes A, B and C, so it should
be called counter-clockwise. :) Sounds reasonable.

In any case, I always thought and I think we agree that node B serves
range B-E, node C serves range C-F, etc. So we had the same assumption
in the most recent discussions.

Ning