The Impact of Network Topology on Performance Metrics and Energy Consumption for Blockchains: Towards Repeatable Benchmarking

As blockchains transition from experimental prototypes to production systems, their evaluation must account not only for performance metrics such as throughput and latency, but also for energy costs as well as the repeatability (same setup, same results) and predictability (stable expectations) of assessment under controlled network conditions. While consensus design is a well-known driver, the role of the underlying network topology is still under-characterized and cloud-based studies are expensive and hard to reproduce.

We present

, a reproducible, cost-aware, and system-agnostic blockchain benchmarking environment that supports both cluster deployments and large-scale emulation, enabling controlled changes of the network topology without altering the application logic. Using , we assess five blockchains (Algorand, Diem, Ethereum Clique, Quorum IBFT, and Solana) across five network topologies (fat-tree, full mesh, hypercube, scale-free, and torus) under realistic workloads spanning transfer transactions and smart-contract execution (DDoS, FIFA, gaming, GAFAM, PayPal, VISA). In addition to average performance and energy, we quantify run-to-run dispersion and use statistical tools (factorial ANOVA and ICC) to attribute variance across factors; we further present a four-level predictability taxonomy (P0-P3) and place our methodology at level P2.

Our results highlight a clear performance-energy trade-off: under intensive workloads, full mesh, hypercube, and torus tend to deliver higher performance, whereas fat-tree and full mesh yield the lowest energy per committed transaction. At the same time, throughput and latency can exhibit extreme worst-case deviations (exceeding 1,200%) in specific blockchain-topology-workload combinations, while energy is comparatively steadier and primarily workload-driven. Overall, Algorand and Diem provide strong energy efficiency with stable mid-range dispersion, Ethereum Clique shows the narrowest run-to-run spread but remains more energy-hungry, and Quorum IBFT and Solana are the most sensitive to topology and load, incurring increasing energy costs with scale under heavy workloads.