In this white paper, we introduce the Thunder protocol, a new and provably secure consensus protocol, which overcomes the two main bottlenecks of classic blockchains:
Throughput: Existing blockchains as employed by major crypto-currencies (e.g., Bitcoin or Ethereum) require flooding the whole network with all the transactions. As a consequence, these blockchains do not scale well to handle a large transaction volume. For instance, Bitcoin and Ethereum can handle less than 10 transactions/second. This small throughput severely hinders a wide-spread adoption of such crypto-currencies. As a comparison, VISA handles 2000 transactions/sec.
Confirmation times: The security of Nakamoto’s blockchain protocol requires the block-time to be significantly larger than the maximum network latency ; this is why in Bitcoin, new blocks of transactions get mined every 10 minutes, and a transaction is only considered confirmed once it has been incorporated in a block and there are 5 subsequent blocks following it; thus it takes on average one hour for a transaction to be confirmed (with low confidence). While Ethereum uses a smaller block-time, the average confirmation time still remains relatively high—approximately 13 minutes. These long confirmation times hinder many important applications (especially smart contract applications).
Our Thunder protocol overcomes both of these issues, achieving both high throughput and fast confirmation time (under typical circumstances within 2 network roundtrips), while being robust up to a 50% attack.
While we are aware of several other recently suggested approaches for overcoming the above two bottlenecks, these approaches are either restricted to only payments (and do not handle smart contracts), or provide security only against a weaker 1/3-attack; additionally, many approaches are based on heuristic arguments, whereas our protocol is accompanied by a rigorous proof of security; see Section 4 for more details.
The design of Thunder is inspired by a new theoretical paradigm for consensus proposed in our earlier work, Thunderella , and provides the first practical instantiation of this general paradigm. This paradigm can be instantiated using either Proof-of-Work or Proof-of-Stake.thunder-whitepaper