Publications

Predictive Placement of Geo-distributed Blockchain Nodes for Performance Guarantee

2024 IEEE 17th International Conference on Cloud Computing (CLOUD), Shenzhen, China, 2024, pp. 66-68, doi: 10.1109/CLOUD62652.2024.00017.

Junseok Lee, Yeonho Yoo, Chuck Yoo, Gyeongsik Yang

Download PDF

Abstract

Blockchain-as-a-service (BaaS) in cloud datacenters is gaining widespread attention due to its high performance and privacy. However, existing BaaS solutions lack a method for deciding the proper placement of blockchain nodes across virtual machines in worldwide datacenters to achieve desired performance. Our motivating experiments show that transaction processing performance (TPS) varies ~31.6% depending on the placements. To provide an automatic placement solution for BaaS, we propose Cyan that predicts the TPS for blockchain node placements. Our evaluations on Google Cloud Platform demonstrate that Cyan improves the TPS guarantee ~2.39x compared to existing techniques.

Paper summary

Place blockchain peers for a throughput target

Predicting throughput before placement can help choose where to run geographically distributed blockchain peers.

Peer placement changes communication delays and blockchain performance. Cyan predicts the throughput of candidate placements and selects the placement closest to a requested throughput using predicted transactions per second.

Learn placement performance before deployment

A dataset generator varies peer counts, placement, transaction characteristics and request rates. Cyan trains a performance predictor on 30,000 records, compares candidate placements using predicted transactions per second, and deploys the selected arrangement. Random forest gives the best prediction accuracy among the evaluated models.

Cyan predicts performance from candidate network configurations and uses those predictions to select a placement.View full-size figure
Cyan predicts performance from candidate network configurations and uses those predictions to select a placement.

Check how often placements meet the target

Experiments use Google Cloud in US-east, Europe-west and Asia-northeast regions. Across 1,000 randomized trials, a placement meets the paper’s guarantee criterion when actual throughput falls within 10% of the desired throughput.

Cyan’s guarantee rate is 2.39× that of random placement and 1.52× that of latency-aware placement. The target is actual throughput within 10% of the requested rate.

Cyan’s rate of attaining throughput within 10% of the desired target, relative to each placement baseline.
BaselineCyan guarantee-rate ratio
Random placement2.39×
Latency-aware placement1.52×

Key takeaways

  1. Blockchain placement must account for the throughput delivered by a network configuration. Predicting the target performance metric gives placement a closer connection to the requested service level.
  2. A placement prediction is useful when it survives deployment. Target-tolerance measurements connect candidate ranking to the throughput actually obtained.