Fractal: A New Paradigm for High-Performance Proof-of-Stake Blockchains

The phenomenal successes of cryptocurrencies like Bitcoin and Ethereum have demonstrated the exciting promise of blockchain in powering a new decentralized economy. Yet, the current technologies exhibit many limitations, from the huge energy waste to the unsatisfying performance. In this talk, we will give a brief overview of the celebrated Bitcoin's proof-of-work (PoW) protocol, and, then, Fractal's completely new blockchain design from networking to consensus layers, to deliver the first large-scale, high-performance, and provably-secure proof-of-stake (PoS) protocol. Our latest benchmark indicates that our cross-layer design can scale to 10,000+ nodes across the globe, and, under a modest bandwidth limit of 20 Mbps, sustains a stellar throughput of 3,000+ tps. The first part of the talk will focus on our new consensus protocol, called ?iChing", a pure PoS protocol that is much more energy- efficient. On one hand, iChing's design closely mimics Bitcoin, the only blockchain that stands the test of time for being secure, decentralized (scale to 10,000+ nodes), and irreversible. On the other hand,"iChing" through rigorous yet practical cryptographic design, addresses notorious security challenges in PoS such as nothing-at- stake and grinding attacks. The second part of the talk will present "BackPacker" our systematical redesign of peer-to-peer network to make permissionless and anonymous communication in the network more robust and fair. Moreover, "BackPacke" presents a new philosophy in separating application data (e.g. transaction information) from consensus data (order of transactions) to substantially reduce the disparity among the participating nodes. Through rigorous proofs and real-world benchmark, we show that "BackPacker" achieves both optimal block propagation time as well as near-optimal throughput (80% of the network-physical-limit). Finally, we will also discuss Fractal's perspective on important topics including effective governance and building decentralized applications. Some results in this talk are based on my joint work with Thang Dinh, Lei Fan, Jonathan Katz, and Phuc Thai.