Long-term data resilience using opinion polls

Abstract

Opinion polls can be used as a means to reach weak agreement, an idea introduced by the LOCKSS system. We investigate a set of protocols based on those of LOCKSS that achieve data resilience for the long-term using a peer-to-peer network, where mutually untrusted peers are loosely organized. Peers use opinion polls to correct corrupted copies of data items instead of conventional methods that use consensus algorithms or cryptography to sign data. We give an overview of how LOCKSS performs opinion polls. We improve the current algorithms and evaluate our protocols in terms of their performance and security against adversary attacks. Finally we investigate the dynamics and steady state of the system.