Towards the Predictability of IPFS Nodes’ Session Time Using Machine Learning

Abstract

The InterPlanetary File System (IPFS) is a representative decentralized data storage system that has been widely used in recent years. IPFS plays an important role in the emerging Web3-related applications. As a global peer-to-peer system, a good understanding of IPFS nodes’ session time is meaningful. In this work, we introduce a measurement study to uncover the issues that are related to the session time of IPFS nodes. Based on the collected massive data of all online IPFS nodes for over one month, we gain a comprehensive understanding of the relationship between various node attributes and session time. In addition, we build a supervised machine learning-based model to predict the session time with a high prediction performance.