.png){kind=icon}
.png){kind=icon}
Return
EbbFlow: Non-blocking join synchronization in dynamic asynchronous BFT
.png){kind=icon}
PRE
AI
DOI：10.1016/j.comnet.2026.112368.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Asynchronous Byzantine Fault Tolerance (BFT) is attractive for blockchains because it preserves safety without relying on timing assumptions. However, existing asynchronous BFT frameworks with dynamic membership still handle replica joins largely through a sequential join path: a newly joining replica must wait until reconfiguration is safely installed before beginning effective state synchronization. This serializes reconfiguration and catch-up, prolongs the join path under churn, and leaves the newcomer idle during a costly part of the transition. This paper presents EbbFlow, a fully asynchronous BFT framework for non-blocking joining. The key idea is to decouple synchronization from voting rights: a new replica first joins as a non-voting observer, validates the current configuration view, and begins synchronization while consensus under the current configuration continues. The joiner becomes eligible to vote only after the next configuration is safely certified and installed. To realize this join path, EbbFlow combines descriptor-guided rateless retrieval, cross-configuration-safe agreement progression, and checkpoint-certified configuration continuity. We prove the safety and liveness of EbbFlow under dynamic membership and evaluate its dynamic performance. The results show that EbbFlow substantially shortens the join-critical path by overlapping synchronization with reconfiguration: at a historical backlog of H=128, it reduces ready-to-vote latency by about 34% compared with Turritopsis. EbbFlow also preserves progress across reconfiguration with transient throughput degradation and incurs modest overhead in the static setting.
Keywords:
Asynchronous Byzantine Fault Tolerance
Dynamic membership
Non-blocking join
State synchronization
Consensus protocols
Journal
IF:
4.6
Papers:
1.5K
Citations:
1.6W