DALedger: Towards High-Performance Transaction Processing for Collaborative Decentralized Applications

Despite recent intensive research, existing blockchain systems still have limitations in supporting decentralized applications. In particular, although existing blockchain systems execute internal transactions of different applications concurrently, it is difficult to process the concurrent control considering both the cross-application transactions and internal transactions. The reason is that each application can only access the transactions related to it, and the conflicts between all transactions cannot be detected by the applications. To improve the concurrency of blockchain systems for decentralized applications, we propose a novel blockchain named as DLedger, which is designed based on the directed acyclic graph ledge structure. DALedger supports the concurrent execution of not only internal transactions in the same application but also internal transactions and cross-DApp transactions. We prove that cycles in the dependency graph must have a special dangerous structure, which can be detected in a single application’s partial dependency graph. Based on that, we propose a novel concurrency control mechanism to resolve concurrency conflicts, while ensuring serializability. We conduct more extensive experiments compared with state-of-the-art blockchain ledgers for decentralized applications. Experimental results show that our method outperforms existing works significantly.