|
The reason UTXOs parallelize better is because when you look at a transaction on a UTXO-based blockchain like Bitcoin, you can deterministically and often immediately know from it exactly how the state of the system will change, and which states will change. A Bitcoin transaction for example completely describes which outputs will be spent, which new outputs will be created, and the new amounts in those outputs. This is a really nice property. It makes validating and analyzing transactions parallelizable by design. The system still has to protect against double-spends globally, but this part is light in comparison and can be very fast. In contrast to Bitcoin, basically every other smart contract blockchain is account-based, not UTXO-based, and has shared global state. This does not parallelize naturally at all. I like to think its similar to using cash vs transferring money using Venmo. Giving someone cash parallelize naturally, like UTXOs, whereas Venmo requires a database lock. In account-based systems, it’s impossible to tell how a transaction will change the system's state without executing it. The order that transactions are executed is crucial too, which is why people say Ethereum is single-threaded. Attempts to parallelize account-based systems fall into different categories. Ethereum 2.0's plan is to basically create separate VMs with separate state, called shards or rollups. This can work sometimes, but it makes interactivity between shards (think applications) much more difficult, and interactivity to me is a major reason for using blockchain. Another approach is to embed information into transactions to make them parallelize more like UTXO-based transactions, which is what Solana does. But they parallelize only at the smart contract level, not at the asset level as UTXOs can. Whereas UTXO-based designs get parallelizability by design, creating a programming model for smart contracts and tokens has to-date been harder on these systems, and frankly there aren't even that many people trying. I've been working on this problem for a while and believe I've found a breakthrough. Check out https://run.network if it interests you. |