LIPIcs.OPODIS.2022.8.pdf
- Filesize: 0.8 MB
- 22 pages
State synchronization, the process by which a new or recovering peer catches up with the state of other operational peers, is critical to the operation of blockchain-based systems. Existing approaches to state synchronization typically involve downloading snapshots of system state. Such approaches introduce an attack vector from malicious peers that can significantly degrade performance. Moreover, the process of creating snapshots leads to performance hiccups. This paper presents a technique for peers to catch up with operational peers without trusting any particular peer and gracefully recover from misbehavior during the process. We have integrated our design into a production blockchain middleware. Our evaluation shows that during operation, the transaction throughput is consistently higher without pauses for snapshot construction. Moreover, the time it takes for a new peer to join the blockchain is halved, while at the same time tolerating Byzantine peers.
Feedback for Dagstuhl Publishing