FAST: A MapReduce Consensus for High Performance Blockchains

Blockchain platforms when used as a database for IoT systems can resolve data reliability, fault-tolerance, consistency and non-repudiation issues. However, their inherent shortcomings related to their throughput in terms of processed transactions, limit their applicability in such environments in a decentralized way as the underlying network is unable to sustain high workloads. In this paper a fully decentralized high performance consensus mechanism, named FAST, is proposed for a public blockchain. FAST is based on mapreduce paradigm for aggregating and adding transactions on blockchain blocks. FAST was implemented and evaluated in a basic blockchain prototype. A light client for FAST using IPFS, was developed to bring about a reduction in the data stored locally. The obtained results from tests conducted on the prototype depict that FAST exceeds the performance of not just other existing blockchain platforms but comes very close to the throughput of traditional electronic payment networks such as Visa.