Scalable Byzantine fault-tolerant storage

Byzantine fault-tolerance is costly, both in terms of increased latency and limited scalability. Most recent contributions in the area have addressed the latency aspect, leaving the throughput scalability problem largely untouched. We propose in this short paper to build scalable Byzantine fault-tolerant storage systems based on the concept of mini-transactions. To achieve this goal we propose a novel atomic commit protocol that tolerates malicious clients and servers.