Resources in a distributed system can be identified using identifiers based on random numbers. When using a distributed hash table to resolve such identifiers to network locations, the straightforward approach is to store the network location directly in the hash table entry associated with an identifier. When a mobile host contains a large number of resources, this requires that all of the associated hash table entries must be updated when its network address changes. We propose an alternative approach where we store a host identifier in the entry associated with a resource identifier and the actual network address of the host in a separate host entry. This can drastically reduce the time required for updating the distributed hash table when a mobile host changes its network address. We also investigate under which circumstances our approach should or should not be used. We evaluate and confirm the usefulness of our approach with experiments run on top of OpenDHT.
Download