SkyHOST: A Unified Architecture for Cross-Cloud Hybrid Object and Stream Transfer

A team of researchers led by Muhammad Arslan Tariq has introduced SkyHOST (Hybrid Object and Stream Transfer), a novel architecture designed to solve a critical pain point in modern multi-cloud data management. Built upon the existing Skyplane framework, SkyHOST provides a unified control plane and command-line interface (CLI) that bridges the traditionally separate worlds of bulk object transfer and continuous data streaming. This means engineers no longer need to deploy and configure distinct tools like Apache Kafka for streams and specialized bulk copiers for large files; SkyHOST handles both paradigms through intelligent, URI-based routing that automatically selects the optimal transfer mechanism.

SkyHOST's key innovation is its ability to manage heterogeneous data movement patterns—from structured records for real-time analytics to large binary objects—under a single operational umbrella. The paper demonstrates its efficacy through an environmental monitoring use case and empirical evaluations, showing that it achieves competitive throughput for cross-region transfers across providers like AWS, Azure, and Google Cloud. By consolidating these workflows, SkyHOST significantly reduces operational complexity, configuration overhead, and the potential for errors that arise from managing multiple, siloed data transfer systems. This represents a meaningful step toward more agile and simplified data infrastructure for organizations running distributed, cloud-native analytics and decision-making pipelines.