Prediction of Permissioned Blockchain Performance for Resource Scaling Configurations

TL;DR

This work tackles the challenge of configuring BaaS for permissioned blockchains by predicting performance under horizontal and vertical scaling in Hyperledger Fabric. It introduces two random-forest models trained on 593 measurements to forecast transaction success rate ($SR$) and throughput ($TPS$) from input features such as the number of peers, CPU quota per peer, and transaction rate, achieving SMAPE values of $5.6\%$ and $1.9\%$, respectively. Two practical use-cases demonstrate the value: (i) selecting cost-efficient scaling configurations that meet SR and TPS targets, and (ii) forecasting maximum throughput for planned demand, with real-system validation showing strong alignment. The results enable proactive, data-driven scaling decisions in BaaS, reducing trial-and-error and supporting reliable operation in private-blockchain deployments, with discussion of generalization to other platforms and future extensions.

Abstract

Blockchain is increasingly offered as blockchain-as-a-service (BaaS) by cloud service providers. However, configuring BaaS appropriately for optimal performance and reliability resorts to try-and-error. A key challenge is that BaaS is often perceived as a ``black-box,'' leading to uncertainties in performance and resource provisioning. Previous studies attempted to address this challenge; however, the impacts of both vertical and horizontal scaling remain elusive. To this end, we present machine learning-based models to predict network reliability and throughput based on scaling configurations. In our evaluation, the models exhibit prediction errors of ~1.9%, which is highly accurate and can be applied in the real-world.

Figures (6)

• Figure 1: HLF components
• Figure 2: HLF transaction flow
• Figure 3: Hyperparameter boundaries and selected values.
• Figure 4: Prediction accuracy comparison
• Figure 5: Predicted performances (500 transaction request rate).