The Impact of Vote Counting Policy on the Performance of PBFT: Invited Paper

Practical Byzantine Fault Tolerance (PBFT) is the protocol of choice for many application that require a distributed agreement or consensus between a number of participants. While PBFT assumes that all participants have an equal say in the final decision, many applications, and recently introduced identity management systems in particular, require that the weights are unequal so that they reflect the trust placed in participants that belong to different categories. To apply PBFT in those scenarios, we propose to introduce a small change to the manner in which the votes are counted in the COMMIT phase. We present the details of two alternative solutions in which orderers' votes are counted separately in two or three groups or committees, and investigate the performance impact of these changes on the mean time to accept a data block and the number of nodes involved in making the final decision. Our results indicate that the proposed solutions impose a slight performance penalty which may be countermanded by reducing the quorum numbers needed in different subsets of the original committee.