Researcher from Red Bay Network presents scientific analysis of blockchain transaction throughput (TPS) claims versus real-world performance, revealing significant gaps between announced metrics and observed results across major blockchains including Algorand, Avalanche, and Solana.
A distributed systems researcher and CEO of Red Bay Network discusses the critical difference between claimed and actual transaction per second (TPS) performance in blockchains. The talk covers why blockchain projects announce inflated TPS numbers, the importance of measuring both throughput and latency together, and introduces Diablo, an open-source benchmarking suite that tests seven blockchains using realistic decentralized applications with real traces from Uber, FIFA, NASDAQ, and YouTube. The research deployed 200 machines across 10 global regions and found that most blockchains failed to commit all transactions under realistic workloads, with performance heavily dependent on experimental setup, hardware resources, and network configuration.