An implementation of consistent hashing ring for distributed hash tables as a gen_server. Here is an informal explanation of the technique: http://www.spiteful.com/2008/03/17/programmers-toolbox-part-3-consistent-hashing/ Exports: start_link(Peers) -> ServerRef start_link(ServerName, Peers) -> ServerRef Types: Peers = [{Node, Opaque, Weight}] Node = term() Opaque = term() Weight = int(), >= 1 Create a ring according to peer configuration Peers. Opaque is an opaque data associated with the node. Greater weight makes the node own bigger portion of the ring. The ring creation time is N^2 where N is the sum of the nodes' weights. add(ServerRef, Peers) -> ok | {error, already_there, [Node]} Add Peers to the ring. If any of the nodes being added is already in the ring (the check is done by comparing node *names*), error is returned and the ring remains intact. add(ServerRef, {Node, Opaque, Weight}) Equivalent to add(ServerRef, [{Node, Opaque, Weight}]). delete(ServerRef, [Node]) -> ok | {error, unknown_nodes, [Node]} Removes nodes from the ring by their names. If any of the names doesn't reference an existing node in the ring, an error is returned and the ring remains intact. delete(ServerRef, Node) Equivalent to delete(ServerRef, [Node]). get_config(ServerRef) -> Peers Returns the current configuration of the ring. The order of nodes within the returned list is not specified. lookup(ServerRef, Key) -> [{Node, Opaque}] Types: Key = term() Look up the Ring for a list of nodes corresponding to Key. The returned list is guaranteed to contain no duplicates. The lookup time is log(N), where N is the sum of the nodes' weights. nodes(ServerRef) -> [{Node, Opaque}] Return the list of nodes in the ring. The order of nodes in the returned list is not specified, but it is guaranteed that the list contains no duplicates. This operation execution time is constant.