FabricTCA: A two-level conflict avoidance approach to consortium blockchains

Abstract

Consortium blockchain systems have been recognized as promising infrastructures for data sharing and collaboration among different organizations due to their trustless nature. However, performance bottlenecks have seriously hindered the applications of these systems in many practical scenarios, especially under high transaction volumes. This study identifies two critical factors affecting system performance: intra-block and inter-block transaction conflicts. To address these issues, we present FabricTCA (Fabric Two-level Conflict Avoidance), a method that significantly improves system throughput by mitigating both transaction-level (intra-block) and block-level (inter-block) conflicts. To resolve intra-block transaction conflicts, we propose a scheduling strategy that detects and prevents conflicts by introducing a novel dependency chain data structure and defining the notion of dangerous structures. For inter-block transaction conflicts, we propose a conflicting transaction detection mechanism by constructing a cache during the early ordering service stage, in which detected conflicting transactions will be sent back to clients and resubmitted. Extensive experiments under multiple workload scenarios demonstrate that FabricTCA outperforms state-of-the-art consortium blockchains, such as Hyperledger Fabric (a.k.a. Fabric) and its multiple variants, in terms of throughput, transaction abort rate, transaction execution time, and space utilization. In particular, FabricTCA achieves a 9.51-fold increase in throughput and reduces space consumption by 88.27 % compared to Hyperledger Fabric, showcasing its superior performance. Additionally, FabricTCA demonstrates low storage occupation, strong robustness to attacks and high detection speed.