Hello guys,

I was looking at the paper : Replication strategies for Reliable
Decentralised Storage

(it can be found at : http://urchin.earth.li/~mleslie/mleslieStorage.pdf)

It describes a strategy : replication at successor, wich is not a real
replication at successor as DHash is, but it tries to imitate the
successor placement with the function :

ReplicaKey(key,repNumber) = key + (replicaNumber * NetSize /numberOfPeers)
mod NetSize.

So, it places replicas of items at regular intervals following the
responsible of the key, in such a way they is a great chance they will be
at successors of the responsible pair.

The question is : It is very strange replica placement is dependent of the
number of peers in the network. What happens if a new node join? We won't
know where are our replicas anymore..because the replica function would
give other values.

Did I miss something? Or perhaps, the maintenance algorithm really
sucks..at each node join, lots of items have to be transfered from lots of
nodes, so that the conditions hold again. This seems really inprobable.

Thank you.

Julien.

_______________________________________________
p2p-hackers mailing list
[email protected]
http://lists.zooko.com/mailman/listinfo/p2p-hackers

Reply via email to