Parallel byzantine fault tolerance

Abstract

The increasing popularity of distributed systems and applications has generated the need for algorithms guaranteeing high availability and reliability of such solutions. As an answer to this demand, various Byzantine fault-tolerant algorithms have been designed, allowing the systems to provide the correct service even in the presence of faults, either accidental or of malicious origin. However, despite significant efforts recently made, their practicality is still limited by many factors, such as the cost of additional machines or decreased system throughput. Addressing these issues, we propose Apex, a parallel Byzantine fault-tolerant execution algorithm. By processing independent requests in parallel on different multicore machines, we were able to obtain a significant increase in throughput over traditional algorithms. The performed tests have shown that our approach can execute the incoming packs of requests even several times faster than other similar algorithms.