[Paper Review] Blockchain Goes Green? Part II: Characterizing the Performance and Cost of Blockchains on the Cloud and at the Edge

Dumitrel Loghin, Tien Tuan Anh Dinh+4 more|arXiv (Cornell University)|May 14, 2022

Blockchain Technology Applications and Security4 citations

TL;DR

This paper evaluates the performance and cost of Hyperledger Fabric and ConsenSys Quorum on diverse hardware—x86/64 and ARM-based cloud and edge systems—demonstrating that ARM-based Amazon Graviton instances outperform Intel Xeon servers in throughput and cost-efficiency, while edge devices offer lower performance but significant long-term cost savings when manpower costs are excluded.

ABSTRACT

While state-of-the-art permissioned blockchains can achieve thousands of transactions per second on commodity hardware with x86/64 architecture, their performance when running on different architectures is not clear. The goal of this work is to characterize the performance and cost of permissioned blockchains on different hardware systems, which is important as diverse application domains are adopting t. To this end, we conduct extensive cost and performance evaluation of two permissioned blockchains, namely Hyperledger Fabric and ConsenSys Quorum, on five different types of hardware covering both x86/64 and ARM architecture, as well as, both cloud and edge computing. The hardware nodes include servers with Intel Xeon CPU, servers with ARM-based Amazon Graviton CPU, and edge devices with ARM-based CPU. Our results reveal a diverse profile of the two blockchains across different settings, demonstrating the impact of hardware choices on the overall performance and cost. We find that Graviton servers outperform Xeon servers in many settings, due to their powerful CPU and high memory bandwidth. Edge devices with ARM architecture, on the other hand, exhibit low performance. When comparing the cloud with the edge, we show that the cost of the latter is much smaller in the long run if manpower cost is not considered.

Motivation & Objective

• To evaluate the performance and cost of permissioned blockchains across diverse hardware architectures, including x86/64 and ARM, in both cloud and edge deployments.
• To address the lack of systematic evaluation of blockchain workloads on non-x86/64 platforms, especially ARM-based systems increasingly used in cloud and edge environments.
• To quantify the total cost of ownership (TCO) of blockchain nodes, including hardware, energy, and manpower costs, for informed deployment decisions.
• To compare the efficiency of two major permissioned blockchains—Hyperledger Fabric and Quorum—across different consensus mechanisms and hardware platforms.
• To guide application designers and system architects in selecting optimal hardware for blockchain workloads based on performance, energy efficiency, and long-term cost.

Proposed method

• Conducted extensive benchmarking of Hyperledger Fabric and Quorum on five hardware platforms: AWS Xeon (x86/64), AWS Graviton (ARM), edge Xeon, NVIDIA Jetson TX2 (ARM), and Raspberry Pi 4 (ARM).
• Measured performance using throughput (transactions per second) and latency (95th percentile) under varying workloads and node counts.
• Quantified total cost of ownership (TCO) using a model that includes hardware acquisition cost, energy consumption (based on average power draw), and manpower costs over a 3-year lifespan.
• Used a standardized TCO formula: $ C = N \cdot C_s + N \cdot T \cdot P_a \cdot C_{ph} + T \cdot C_{mh} $, where $ N $ is number of nodes, $ C_s $ is node cost, $ T $ is lifespan in hours, $ P_a $ is average power, $ C_{ph} $ is electricity cost per kWh, and $ C_{mh} $ is manpower cost per hour.
• Evaluated both Raft (CFT) and IBFT (BFT) consensus in Quorum to represent different fault-tolerance requirements.
• Compared cloud (IaaS) and edge deployments under two scenarios: with and without manpower cost, to assess operational sustainability.

Experimental results

Research questions

• RQ1 How do performance and cost of permissioned blockchains vary across x86/64 and ARM-based hardware in cloud and edge environments?
• RQ2 What is the relative performance of Hyperledger Fabric and Quorum on ARM-based Graviton instances compared to x86/64 Xeon servers?
• RQ3 How does the total cost of ownership (TCO) of blockchain nodes differ between cloud and edge deployments when excluding and including manpower costs?
• RQ4 To what extent do edge devices with ARM CPUs offer cost advantages despite lower throughput?
• RQ5 How do different consensus mechanisms (Raft vs. IBFT) in Quorum affect performance and cost across hardware platforms?

Key findings

• ARM-based Amazon Graviton cloud instances achieved up to 10% higher throughput than Xeon-based instances for Hyperledger Fabric and were 35% cheaper, primarily due to higher CPU performance and 2× greater memory bandwidth.
• For Quorum, Graviton instances delivered 25% lower throughput than Xeon instances but still offered a 35% cost reduction, indicating better cost-efficiency despite lower performance.
• Edge devices such as Raspberry Pi 4 and Jetson TX2 exhibited significantly lower performance—up to 10× slower than cloud instances—due to limited CPU and memory resources.
• When manpower costs were excluded, edge deployments were substantially cheaper: Xeon(edge) was 1.5× cheaper, Jetson TX2 10× cheaper, and Raspberry Pi 4 75× cheaper than Graviton cloud instances.
• Including manpower costs reversed the cost advantage of edge, making cloud deployments 2× cheaper per hour on average (e.g., $0.476/hour for Xeon(cloud) vs. $0.884/hour for Xeon(edge)).
• The results indicate that long-term cost savings from edge deployments are highly dependent on whether operational manpower is factored into TCO, suggesting that edge may be more viable for low-maintenance, low-throughput applications.