What Blocks My Blockchain's Throughput? Developing a Generalizable Approach for Identifying Bottlenecks in Permissioned Blockchains
Orestis Papageorgiou, Lasse Börtzler, Egor Ermolaev, Jyoti Kumari, Johannes Sedlmeir
TL;DR
This paper presents a generalized, graphical approach to identifying bottlenecks in permissioned blockchains by combining a systematic literature review with end-to-end DLPS benchmarking and enhanced visualization. The method is demonstrated through in-depth studies of Hyperledger Fabric and Quorum, using a two-stage analysis that first detects candidate bottlenecks from node-resource metrics and then correlates them with throughput across multiple time scales. The authors find that Fabric's throughput is CPU-bound within the peer validation phase (vscc/mvcc), while Quorum's performance is constrained by executable-transaction-pool limits and limited parallelism, leading to significant rejections before reaching maximum load. The work provides researchers and practitioners with a practical toolkit, data-documentation guidelines, and a structured search workflow for bottlenecks, contributing to more reproducible and comparable evaluations of permissioned blockchain performance.
Abstract
Permissioned blockchains have been proposed for a variety of use cases that require decentralization yet address enterprise requirements that permissionless blockchains to date cannot satisfy -- particularly in terms of performance. However, popular permissioned blockchains still exhibit a relatively low maximum throughput in comparison to established centralized systems. Consequently, researchers have conducted several benchmarking studies on different permissioned blockchains to identify their limitations and -- in some cases -- their bottlenecks in an attempt to find avenues for improvement. Yet, these approaches are highly heterogeneous, difficult to compare, and require a high level of expertise in the implementation of the underlying specific blockchain. In this paper, we develop a more unified and graphical approach for identifying bottlenecks in permissioned blockchains based on a systematic review of related work, experiments with the Distributed Ledger Performance Scan (DLPS), and an extension of its graphical evaluation functionalities. We conduct in-depth case studies on Hyperledger Fabric and Quorum, two widely used permissioned blockchains with distinct architectural designs, demonstrating the adaptability of our framework across different blockchains. We provide researchers and practitioners working on evaluating or improving permissioned blockchains with a toolkit, guidelines on what data to document, and insights on how to proceed in the search process for bottlenecks.