Dependency-Aware Execution Mechanism in Hyperledger Fabric Architecture

2025-09-09

AI Summary

Hyperledger Fabric suffers from low throughput and high transaction rejection rates under high load, primarily due to serial bottlenecks across endorsement, ordering, and validation phases, as well as late conflict detection and resource contention induced by optimistic concurrency control. This paper proposes a dependency-aware execution mechanism: (1) marking read/write-set dependencies via hash-based mapping during endorsement; (2) enhancing the ordering service to construct dependency-annotated blocks; and (3) embedding an intra-block directed acyclic graph (DAG) to explicitly encode transaction dependencies, enabling DAG-guided parallel execution at commit nodes. Our approach is the first to tightly integrate dependency-aware scheduling with block-level DAG structure, preserving both parallelism for independent transactions and strict ordering for dependent ones. Experiments demonstrate up to a 40% throughput improvement and significantly reduced rejection rates under high contention, while maintaining full compatibility with Fabric’s existing consensus and smart contract layers.

Technology Category

• Search and Optimization: Distributed Search
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems
• Machine Learning: Hardware-aware ML

Application Category

• Security and Privacy: Blockchains and distributed ledgers
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Decentralized Web and Fediverse systems

Abstract

Hyperledger Fabric is a leading permissioned blockchain framework for enterprise use, known for its modular design and privacy features. While it strongly supports configurable consensus and access control, Fabric can face challenges in achieving high transaction throughput and low rejection rates under heavy workloads. These performance limitations are often attributed to endorsement, ordering, and validation bottlenecks. Further, optimistic concurrency control and deferred validation in Fabric may lead to resource inefficiencies and contention, as conflicting transactions are identified only during the commit phase. To address these challenges, we propose a dependency-aware execution model for Hyperledger Fabric. Our approach includes: (a) a dependency flagging system during endorsement, marking transactions as independent or dependent using a hashmap; (b) an optimized block construction in the ordering service that prioritizes independent transactions; (c) the incorporation of a Directed Acyclic Graph (DAG) within each block to represent dependencies; and (d) parallel execution of independent transactions at the committer, with dependent transactions processed according to DAG order. Incorporated in Hyperledger Fabric v2.5, our framework was tested on workloads with varying dependency levels and system loads. Results show up to 40% higher throughput and significantly reduced rejection rates in high-contention scenarios. This demonstrates that dependency-aware scheduling and DAG-based execution can substantially enhance Fabric's scalability while remaining compatible with its existing consensus and smart contract layers.

Problem

• Improving transaction throughput in Hyperledger Fabric under heavy workloads
• Reducing transaction rejection rates caused by validation bottlenecks
• Addressing resource inefficiencies from optimistic concurrency control mechanisms

Innovation

• Dependency flagging system using hashmap during endorsement
• Optimized block construction prioritizing independent transactions
• DAG-based parallel execution for dependent transaction ordering