[chord] Why a key is stored at its successor node?

[Emil Sit]

On Wed, 16 April 2008 at 20:55 (+0200), Manuel Urueña wrote:
> Therefore, if the key is stored at the predeccessor node instead of the 
> succesor one, the last routing hop between the predeccessor and the successor 
> nodes could be saved.

The routing algorithm actually terminates at the predecessor: the
predecessor returns the successor to the query initiator.

Various performance optimizations are described in our NSDI 2004 paper.

> On the other hand, if the key is replicated in some predeccessor nodes, these 
> copies could be found earlier, during the routing process, lowering the 
> average number of hops, and also providing load balancing among replicas.

The SOSP 2001 paper does discuss replicating keys along the lookup path.
See Section 5.2.  For an alternate take on this, see the Beehive paper.
http://www.cs.cornell.edu/people/egs/beehive/

-- 
Emil Sit / MIT CSAIL PDOS / http://pdos.csail.mit.edu/chord/