Consensus based High Consistency Ledger Pruning for DAG-based IoT Blockchain

Abstract

Directed Acyclic Graph (DAG) based Distributed Ledger Technologies (DLTs) are promising solution for IoT environments, offering scalability and high transaction throughput. Given the resource constraints of IoT Gateways (GWs), which maintain the ledgers, pruning of a ledger is important to prevent its size from increasing beyond the storage capacity of GWs. However, existing methods cannot simultaneously address the scalability and consistency issues for planning and executing the DAG based ledger pruning, consistency preservation, and orphaned transactions. We explore an approach that Storage Nodes (SNs) of the DAG based DLT plan the ledger pruning in an efficient, scalable, and consistent manner, and then GWs execute it. This paper introduces the Consensus based High Consistency Ledger Pruning (CHCLP) Framework, which uses two key techniques: Pruning List Reduction (PLR) and Consensus-based Pruning Mechanism (CPM). PLR aggressively compresses a long list of prunable transactions to the combination of a single transaction called "Adam Point" and the rest encoded as unique bit string paths from the Adam Point. CPM leverages Practical Byzantine Fault Tolerance consensus among SNs to ensure tamper-resistant agreement on the Pruning List (PL) and maintain DAG consistency across GWs. The CHCLP Framework was implemented by extending an existing benchmark DAG based DLT system. The proposed framework reduced the size of a PL by 98%, and the ratio of orphan transactions decreased to 0.5% compared to the benchmark system, which was 10.3%. The CHCLP Framework reduces the storage consumption of GWs in DAG based DLTs to a desirable extent with high consistency, and can drastically improve their scalability and performance.