Authored by:
Pouriya Zarbafian, University of Sydney
Vincent Gramoli, University of Sydney, CTO Redbelly Network

In State Machine Replication (SMR), preventing reordering attacks by ensuring a high degree of fairness when ordering commands requires that clients broadcast their commands to all processes. This is impractical due to the impact on scalability, and thus it discourages the adoption of a fair ordering of commands. Alternative approaches to order-fairness allow clients to send their commands to only one process, but provide a weaker notion of order-fairness. In particular, they disadvantage isolated processes. In this paper we introduce Aion, a set of order-fair protocols for SMR. We first leverage trusted execution environments (TEEs) to enable processes to compute the times when commands are broadcast by their issuers. We then integrate this information into existing consensus protocols to devise order-fair SMR protocols that are both leader-based and leaderless. To realise order-fairness, Aion only requires that a client sends its commands to a single process, while at the same time enabling precise ordering during synchronous periods.