A storage-efficient learned indexing for blockchain systems using a sliding window search enhanced online gradient descent

Asiamah, Emmanuel Acheampong, Akrasi-Mensah, Nana Kwadwo, Odame, Prince, Keelson, Eliel, Agbemenu, Andrew Selasi, Tchao, Eric Tutu, Al-Khalidi, Mohammed ORCID: https://orcid.org/0000-0002-1655-8514 and Klogo, Griffith Selorm (2025) A storage-efficient learned indexing for blockchain systems using a sliding window search enhanced online gradient descent. The Journal of Supercomputing, 81 (1). 321. ISSN 0920-8542

Preview

Accepted Version Available under License Creative Commons Attribution. Download (2MB) | Preview

Abstract

With its promise of transparency, security, and decentralization, blockchain technology faces significant challenges related to data storage and query efficiency. Current indexing methods, which often rely on structures like Merkle trees and Patricia tries, contribute to excessive storage overhead and slower query responses, particularly for full nodes that maintain a complete copy of the blockchain. To address this, we introduce a novel-learned indexing approach for blockchain that utilizes a layered structure with a sliding window search enhanced Online Gradient Descent (SWS-OGD) as the inter-block index. The method was implemented across five distinct blockchain environments—Bitcoin, Ethereum, Dogecoin, Litecoin, and IoTeX. Experimental results demonstrate that the proposed method reduces storage costs by up to 99% compared to state-of-the-art approaches, requiring as little as 0.9 KB for 20,000 blocks-a substantial improvement over existing methods. Despite the significant reduction in storage costs, the SWS-OGD method maintains comparable performance in other key metrics, such as query latency. These results ensure that blockchain systems can handle large-scale data queries efficiently, maintaining high performance even as the blockchain grows in size.